Towards resolving the double classification in Erythraeus (Actinotrichida: Erythraeidae): matching larvae with adults using 28S sequence data and experimental rearing

The taxonomy of free-living adults and heteromorphic parasitic larvae of Parasitengona mites has in the past been treated independently resulting in a double classification. Correct linkage of names still remains unknown for many species. A holistic understanding of species is imperative for understanding their role in ecosystems. This is particularly true for groups like parasitengone mites with a radically altered lifestyle during development—parasitic to predatory. Here, we infer linkages of three nominal species of Erythraeus, using matching with 28S DNA sequence data from field-collected specimens and through laboratory rearing. The general mixed Yule coalescent method (GMYC) was used to explicitly test if field-collected specimens representing heteromorphic life instars were conspecific. The field-collected larvae were allocated to adults of Erythraeus cinereus and Erythraeus regalis, respectively. Laboratory rearing of the same two species confirmed the matching done by DNA. Rearing was also successful for Erythraeus phalangoides after eggs were treated to an imitated winter diapause. This integrative taxonomic approach of molecular, morphological, and rearing data resulted in the following synonyms: E. phalangoides (De Geer, 1778) [= Erythraeus adrastus (Southcott, 1961), syn. nov.], E. cinereus (Dugès, 1834) [= Erythraeus jowitae Haitlinger, 1987, syn. nov.], and E. regalis (C.L. Koch, 1837) [= Erythraeus kuyperi (Oudemans, 1910), syn. nov., = Erythraeus gertrudae Haitlinger, 1987, syn. nov.]. The molecular evidence confirmed the separate identity of three further members of the genus. We provide redescriptions of E. phalangoides, E. cinereus, and E. regalis after modern standards, and neotypes are designated.


Introduction
Few organisms look like small exact copies of the sexually mature adult already at birth, that is, some changes other than just isometric growth are commonly associated with the development from juvenile to adult organism. With great change, e.g., through metamorphosis, species-level identification will rely on completely different sets of morphological or anatomical characters for different life history forms. For the huge diversity of arthropods, with vast numbers still to be discovered and described (Erwin 1982;Basset et al. 2012), it is not surprising that most are known exclusively from a single life history form. For arthropods in general, this is commonly the adult but for various groups it can be the larval form (e.g., vertebrate parasitizing Trombiculidae). Connecting a species' life history forms, especially when biology changes substantially through development, is of course immensely important for understanding a species function in its ecosystem. As an example, malaria control efforts would look very different today if the life history forms of its vector were not connected. A particular problem arises if separate scientific schools, or scientists during different time periods, name life history forms in isolation of each other's work. This results in a Bdouble classification^-many of the separate adult and larval names must be conspecific since they are based on the same regional fauna but to infer the correct linkage is non-trivial (Łaydanowicz and Mąkol 2010). Such is the current situation for erythraeid mites where in the past the taxonomy was largely based on adults but more recently has been based on the parasitic larval forms (Mąkol and Wohltmann 2012).
Until recently, the only available method to match larvae and postlarval forms of parasitengone mites was experimental rearing of field-collected ovigerous females or of parasitic larvae (Southcott 1961). Because of sampling effort, captivity stress, and the complexity of factors influencing parasitengone life cycles, such attempts have been rarely performed for Erythraeidae in the past (Mąkol andWohltmann 2012, 2013).
Molecular methods for species identification now offer alternative ways to both study trophic interactions and to correlate life history forms of species (see Kress et al. 2015 for a recent review). Since the seminal paper by Hebert et al. (2003), this is broadly referred to as DNA barcoding, although in its strict sense, this term only applies to when certain standard genetic markers are being used (mitochondrial COI in animals). Larvaladult matching using DNA has been successfully applied to fish (Hubert et al. 2010;Baldwin and Johnson 2014 and references therein), amphibians (Vences et al. 2005), various invertebrate groups (Jousson et al. 1999;Locke et al. 2011;Okassa et al. 2012;Alcántar-Escalera et al. 2013) including crustaceans (Feller et al. 2013), and a range of insect groups such as beetles (Miller et al. 2005;Caterino and Tishechkin 2006;Ahrens et al. 2007; Levkanicova and Bocak 2009;Lefort et al. 2012), Lepidoptera (Prado et al. 2011), Diptera (Pfenninger et al. 2007;Sutou et al. 2011), caddisflies (Johanson 2007;Zhou et al. 2007;Waringer et al. 2008), mayflies (Gattolliat and Monaghan 2010), stoneflies (Mynott et al. 2011;Avelino-Capistrano et al. 2014), and hemipterans (Zhang et al. 2008;Zhang and Weirauch 2011) to name a few. The main challenge is the unavailability of reference sequences of adults (Pfenninger et al. 2007) or insufficient field sampling of both life instars. Limited amount of DNA that can be obtained from relatively small specimens requires DNA extraction from the entire specimen. However, non-destructive methods of DNA extraction make it possible to retain the exoskeleton for voucher slide preparation (Cruickshank 2002;Dabert et al. 2008). Another challenge can be that the selected molecular marker is not variable enough for species-level identification. This can be due to that the marker itself is conservative with a slow substitution rate or that the species group is relatively young (Hickerson et al. 2006;Bergsten et al. 2012). Mitochondrial genes, like the animal COI barcode, can also portray false species identifications (and false adult-larval linkages) due to hybridization among species and introgression (Whitworth et al. 2007). Despite these challenges, and along with the relatively low cost and time expense, the technique offers an alternative to rearing and is indeed one of the most useful applications of DNA barcoding (Hebert and Gregory 2005;Kress et al. 2015).
In the present study, we test the validity of different Erythraeus species described independently as larvae and adult instars. Two separate methods are used to increase the accuracy in adult-larval linkage and to evaluate the applicability of molecular analysis compared with experimental rearing in terrestrial parasitengone mites. As a consequence of the results, we provide redescriptions for Erythraeus phalangoides (De Geer, 1778), Erythraeus cinereus (Dugès, 1834), and Erythraeus regalis (C.L. Koch, 1837), along with a list of new junior synonyms and neotype designations.

Material and methods
Collecting, mounting, and morphological studies Larvae and adults of Erythraeus spp. were collected at different localities throughout Sweden, Germany and Poland. Data Populations from which specimens were selected for molecular analyses (number of specimens) on localities, sampling time, method and frequency of sampling, collectors' name, and taxon ID, along with the data on material that underwent successful rearing and material dedicated for molecular studies, including Curteria episcopalis (C.L. Koch, 1837) selected as an outgroup, are provided in Table 1. For morphological analyses, larvae and adults selected randomly from each sample period were used. Alcoholpreserved, non-extracted specimens were cleared in potassium hydroxide or in lactic acid prior to mounting. The orcein dye was used in order to stain the exoskeletons of DNA-extracted larvae. The material was mounted on permanent slides in polyvinyl-lactophenol (material from Germany), euparal (material from Sweden), or in Faure's medium (material from Poland). Measurements and drawings were made using Nikon Eclipse 80i, Nikon Eclipse E600, or BB-2^Olympus microscope, equipped with differential interference contrast (DIC) and drawing tube. All measurements are given in micrometers (μm). The total length of legs includes coxae. Terminology, abbreviations applied to morphological structures, and measurement standards follow Wohltmann et al. (2007) and Mąkol (2010), with the following modifications: N = non-specialized, nude seta (may have minute setules) on palps and legs and B = non-specialized, barbed, or setulated seta on palps and legs. Data on deutonymphs, due to their overall similarity to adults but also due to unrecognized intraspecific variation of metric and meristic data, apply only to the life instar behavior.
As the type material of E. phalangoides, E. cinereus, and E. regalis was unavailable for studies, we referred to original descriptions and complementary data published in an array of papers (De Geer 1778;Dugès 1834;Koch 1837;Oudemans 1910Oudemans , 1929Southcott 1961;Haitlinger 1987;Wohltmann et al. 2007), supplemented with reference material from authors' collections. Three further members of the genus (Erythraeus sp. 1, E. sp. 2, E. sp. 3) for which the separate identity was confirmed by means of molecular analyses were considered in the present study in order to strengthen the evidence on the applicability of molecular methods in separation of specific level taxa. The comprehensive characteristics of these species will constitute the subject of a separate study. DNA extractions are stored at the Swedish Museum of Natural History (NHRS), whereas voucher specimens, apart from neotypes (see: type material in systematic part of BResults and discussion^), are at the author's collections. Sequences (E. phalangoides, E. cinereus, and E. regalis) are deposited in GenBank under accession numbers KU892288-KU892385.

Molecular analyses
A non-destructive method of DNA isolation was applied to specimens used for molecular analysis (Table 1), which allowed the exoskeletons to be slide mounted post-extraction (Cruickshank 2002;Dabert et al. 2008). The molecular work was carried out at the Molecular Systematics Laboratory (MSL), Swedish Museum of Natural History, Stockholm. Four legs from one side of mite's body or the whole mite, depending on the size of specimen, were subject to extraction. The extraction method followed the Cell and Tissue DNA Kit protocol of the KingFisher™ Duo (Thermo Scientific). Lysis (56°C) of larvae included 3-4 h and 15-16 h (overnight) for adults. The D2 region of nuclear 28S rDNA (28S) was amplified using primers D2F forward 5′-AGTCGTGTTGCTTGATAGTGCAG-3′ and D2R rev e r s e 5 ′ -T T G G T C C G T G T T T C A A G A C G G G -3 ′ (Campbell et al. 1993;Goolsby et al. 2006). PCR cycle and sequencing protocols followed Stålstedt et al. (2013).
The primers amplifying the D2 region of 28S gave a sequence length of 582-587 bp with 116 variable characters, of which 51 were parsimony informative. Sequences with more than 10 % missing data were excluded. The average nucleotide composition were A = 28 %, G = 27 %, T = 26 %, and C = 19 %.
Raw sequence data were assembled using Sequencher v. 4.10.1 (2008). Primer sequences were removed from the beginning and end of each sequence. The 28S sequences were aligned using FFT-NS-i in MAFFT v.7 (Katoh and Standley 2013) which resulted in an alignment length of 591 bp. In total, 98 samples, including C. episcopalis (C.L. Koch, 1837), taxon included as an outgroup, were used to reconstruct a gene tree with Bayesian analysis, using MrBayes v.3.2.4 (Ronquist et al. 2012). The substitution model GTR + Γ was selected with MrModeltest v. 2.3 (Nylander 2004). The prior on branch lengths was set to an exponential distribution with a mean of 0.01 to prevent the exploration of artificially long trees (Brown et al. 2010). Markov chain Monte Carlo (MCMC) ran for 20 million generations, sampled every 500 generations and with the first 25 % of samples discarded as burn-in. The standard deviation of split frequencies between runs was below 0.01 (=0.0016). The generalized mixed Yule coalescent model (GMYC) (Pons et al. 2006;Fujisawa and Barraclough 2013) was used as a method of molecular species delimitation on the single-locus gene tree. The GMYC analysis requires an ultrametric tree which was inferred with a clock model in BEAST v.2.1.3 (Bouckaert et al. 2014). Duplicate haplotypes were removed from the matrix using DNAcollapser (Villesen 2007) since identical sequences can cause problems with the GMYC model (Fujisawa and Barraclough 2013). The outgroup was excluded as the clock model roots the tree. The analysis contained 45 haplotypes and the settings were GTR + Γ, strict clock, coalescent constant population tree prior, and 20 million generations (MCMC) with 25 % burn-in. The GMYC analysis was performed in R version 3.1.3 with the Bsplitsp ackage (Ezard et al. 2014; R Core Team 2015). Nucleotide composition statistics, genetic intra-and interspecific distances (Kimura 1980), and parsimony informative characters were obtained using MEGA v.6.06 (Tamura et al. 2013).

Experimental rearing
Experimental rearing aimed at obtaining larvae from field-collected females and recognition of the diapausing instar of particular species. Live specimens were kept in laboratory in polystyrene boxes (25 × 25 × 20 mm) (material from Germany) or glass vials (24 × 34 mm) covered with semi-transparent lid (material from Poland) and filled to ca 2/3 with charcoaled Plaster-of-Paris. The material was kept at controlled temperature (20°C, ±1°C tolerance) and light (12 h dark, 12 h light) in environmental chambers. In case of successful development of eggs into larvae, the recorded rearing success was >90 %. When development stopped at a certain instar, specimens were exposed to 5°C for about 90 days and subsequently re-exposed to 20°C in order to imitate winter conditions and to break diapause. Saturated air humidity and humidity of the substratum (indicated by changes in color of charcoaled Plaster-of-Paris) were attained by the addition of water to the vial. Test on the influence of humidity followed Fig. 1 Majority-rule consensus from the non-clock Bayesian phylogenetic analysis of 28S from Erythraeus larvae and postlarval instars sampled in the field. Numbers above branches denote posterior probability values. The clusters of Erythraeus cinereus, E. regalis, E. sp. 1, and E. sp. 2 have both larvae (gray) and adults (black) included. Curteria episcopalis was used as outgroup to root the tree Red lines (thinner lines on the right side of the graph) indicate within-species branches, and black lines (thicker lines on the left side of the graph) represent between-species branches as inferred by the GMYC model procedures described in Wohltmann (1998). Aphididae were offered as hosts to all three Erythraeus species, while larvae of E. cinereus were also given the immature forms of Formicidae as well as deutonymphs and adults of Miridae. Active postlarval forms of E. cinereus and E. regalis were offered larvae and pupae of Brachycera (Diptera) and Formicidae (Hymenoptera) as prey.

Results and discussion
Field data Erythraeus spp. was sampled at various locations in Sweden, Poland, and Germany (Table 1). Seasonal abundance of instars (Table 2) did not differ within a particular species sampled at various localities. Postlarval mobile instars were active on soil surface and in the litter layer. Larvae were collected either as free-living, during host search, or attached to aphid host during parasitic phase (Table 2). Only few of the immobile instars were collected in crevices. Edaphic occurrence as reported for other Parasitengona (Wohltmann 2000 and references therein) was not observed in any of the Erythraeus spp. In case of syntopic occurrence of Erythraeus sp. (10 % of studied cases-see Table 1), species usually differed in the annual abundance of instars. Data on habitat preferences and details concerning life history patterns of Erythraeus spp. are provided in systematic part of BResults and discussion^and in Tables 1 and 2. Patterns concerning phenology and biology of Erythraeus spp., observed during present study, confirm earlier reports on European members of the genus (Beron 1982;Wohltmann 2000Wohltmann , 2001 and fit also general data provided by Southcott (1961Southcott ( , 1988.

Molecular data
The gene tree from the Bayesian analysis grouped the sequences into five distinct clusters and one singleton (Fig. 1). Four clusters contained sequences from both larvae and adults (n = 87) (Fig. 1). All clusters had strong support (>0.95 posterior probability). The GMYC model delimited six to nine entities within a 2 log likelihood unit confidence interval. The maximum likelihood solution was six entities and had a significantly better fit than the null model of a single species (logL = 400.5 versus 383.4, df = 2, p value = 0.0008 The GMYC model of molecular species delimitation analytically confirmed the visual intuition of clusters from the Bayesian analysis as separate species (Fig. 2). The genetic distances between larvae and adults here inferred to be conspecific were also very low (0-0.5 %). In contrast, the lowest interspecific distance was 2.8 %. In comparison to the mitochondrial barcoding gene COI, nuclear D2 region of 28S is a more conservative gene region and is expected to show less variation (Stålstedt et al. 2013). The inter-cluster distances are therefore too high to reflect within-species variation. Levels of D2 28S differentiation between species were lower than observed in aquatic Parasitengona of the genera Hygrobates C.L. Koch, 1837, Piona C.L. Koch, 1842, and Unionicola Haldeman, 1842(Martin et al. 2010Stålstedt et al. 2013) but similar to those recorded for the majority of currently recognized congeneric species, including other actinotrichid mites (Skoracka and Dabert 2010;Skoracka et al. 2013). The intraspecific variation was within the range found for other Parasitengona species (Stålstedt et al. 2013).

Rearing data
Successful laboratory rearing was carried out for each of the three Erythraeus spp. sampled in the field (Tables 1 and 2). Eggs were ovoid and orange just after deposition but turned black within 24 h. Larvae hatched about a month later in E. cinereus and E. regalis, while eggs of E. phalangoides underwent obligatory diapause and developed into prelarvae and subsequently into larvae only when kept at 5°C for the period amounting to about 6-10 months, followed by reexposure to 20°C. After hatching, larvae displayed positive phototactic response. Unfed larvae survived without hosts for a few days only (Table 2). When kept together with aphids, larvae readily started parasitism. Fully engorged larvae moved comparably fast but did not display positive phototactic response. Instead, they tried to hide in crevices, where they became motionless and entered the calyptostatic protonymph. Deutonymphs and adults were fed with larvae and pupae of brachyceran flies and ants. In E. regalis, the deutonymphs displayed obligatory diapause and continued development into subsequent instars only after chilling at 5°C for about 100 days. For E. cinereus, the obligatory diapause was recorded in adults, before the onset of reproduction. Additional laboratory data on E. phalangoides, E. cinereus, and E. regalis are provided under respective species description and in Table 2. Data on biology and life history patterns confirm earlier reports by Wendt (1996) and Wohltmann (2000Wohltmann ( , 2001, applying to Erythraeus spp.

Systematic part
Two-way linkage of larvae and postlarval forms-DNA analysis of field-collected specimens and/or laboratory rearing of Fig. 6 Erythraeus phalangoides, larva. a Gnathosoma and idiosoma, dorsal view. b Dorsal opisthosomal setae. c Gnathosoma and idiosoma, ventral view. d Seta ps larvae from ovigerous females-resulted in the following synonyms: E. phalangoides (De Geer, 1778) [= E. adrastus (Southcott, 1961), syn. nov.], E. cinereus (Dugès, 1834) [= E. jowitae Haitlinger, 1987, syn. nov.], and E. regalis (C.L. Koch, 1837) [= E. kuyperi (Oudemans, 1910), syn. nov., = E. gertrudae Haitlinger, 1987, syn. nov.]. The following redescriptions include variation of morphological and life history traits between different populations (Sweden, Poland, and Germany) of the particular species. In order to maintain the stability of nomenclature, we are of the opinion that the names of species described in the eighteenth and nineteenth century with an insufficient set of characters in the original descriptions should nevertheless be retained due to their extensive usage and, since they do not contradict the present knowledge of the taxa in question, be supported by much more extended characteristics.
Biology Field data and laboratory data are provided in Table 2. E. phalangoides is associated with a wide range of habitats, from sandy dunes close to marine coasts over forested biotopes in lowlands to xeric meadows. In late May to July, deutonymphs and adults were observed to move quickly on the surface of litter layer and to feed on adult ants. Moreover, a single case of preying of adult E. phalangoides on a small beetle was observed. Larvae parasitized aphids in the field and in the laboratory.

E. cinereus
Biology Field data and laboratory data are provided in Table 2. E. cinereus was collected at forests, forest margins, at flooded valleys and the reed belts on the ponds, at one locality part of the habitat was exposed to annual flooding by a nearby creek in winter and in spring. The maximum abundance of larvae was observed in July. Deutonymphs were found in litter layer. Adults were active at sunny days in August and September but remained motionless preferably at the underside of rotting leaves when temperatures were below 10°C. In January and February, diapausing adults were sometimes found frozen but became active when exposed to 20°C in the laboratory. In the field, adults became active again in spring, when they searched the surface of the litter layer for prey and mates.
In the laboratory, only adults collected in spring deposited eggs (n = 31), while females collected in autumn or winter never did so (n = 40). E. cinereus eggs remained viable at 76-100 % relative air humidity and survived even permanent submersion in tap water. Partially fed larvae became attractive as host to unfed conspecifics, which in turn started parasitism. These larvae could again become attractive to other unfed larvae, and in several cases, chains of up to four E. cinereus larvae parasitizing each other were found, which votes for the phenomenon of hyperparasitism existing in Erythraeus spp. In a few cases, both the larva parasitizing the aphid host as well as another larva parasitizing its conspecific reached the next life instar. However, in majority of cases, the uptake of nutrients from conspecifics was not sufficient and, after detachment, larvae continued to display host searching behavior, sometimes leaving larvae dead and sucked empty. Parasitism on unfed larvae of the same species was observed exceptionally (n = 3), 7 days after hatching, and none of the parasitic larvae reached the subsequent life instar. Larvae of ants, nymphs, and adults of plant bugs were not accepted by E. cinereus larvae as potential hosts. Deutonymphs and adults fed on larvae and pupae of ants and brachyceran flies. Adults kept at constant 20°C moved around and fed on prey; however, neither spermatophores nor eggs were deposited. Exposure to 5°C for 71 days again did not induce reproductive behavior after re-exposition to 20°C (n = 15). However, when chilled for 105 days, males deposited spermatophores 5-7 days after re-exposition to 20°C (n = 3), which points at adult as an hibernating instar.
Biology Field data and laboratory data are provided in Table 2. E. regalis was collected close to water reservoirs, both in meadows and in forested areas, however not at inundated places. A single protonymph was collected in the litter layer in August. Deutonymphs were active at sunny days in August-September and again in March moved around on the litter layer. Outside the time of increased activity, they hid in crevices of the bark of Betula 1-3 m above ground level. In May 16, tritonymphs were found in crevices, most in the bark of birch trees and only few in the litter layer. Adults were active in May-June. In the laboratory, eggs were deposited by females collected in early summer. Larvae hatched a month later and survived for up to 13 days without hosts (n = 60, 20°C). When exposed to Aphis sp., larvae attached to the host at first contact. Further development stopped when reaching the deutonymph instar. Although deutonymphs moved and fed on larvae and pupae of ants and brachyceran flies, none entered the tritonymph when kept at constant 20°C. However, of ten deutonymphs captured in September and chilled at 5°C and darkness for 102 days, one entered the tritonymph when subsequently exposed to 20°C and 16/8 light/darkness. The latter constitutes an evidence for deutonymph being a diapausing and hibernating instar.

Concluding remarks
For E. regalis and E. cinereus, the results of the molecular analysis were confirmed by laboratory rearing. In case of E. phalangoides, matching of larvae and active postlarval forms was based on the results of experimental rearing, whereas the distinct ID of the species was confirmed by molecular data. Both molecular analyses and experimental rearing allowed for unambiguous correlation of heteromorphic instars and, in consequence, subsequent morphological investigations of specimens collected at different geographical areas.
The correlation of heteromorphic, metamorphosed instars through DNA analysis and laboratory rearing has been infrequently applied up to date in relation to various non-parasitengone arthropods (Grutter et al. 2000;Barros-Battesti et al. 2007;Kumar et al. 2007;González et al. 2015). As application of both methods led to the same results, we are convinced that DNA barcoding is a highly cost-effective alternative or complement to experimental rearing practice, especially when it comes to problematic species (Hebert et al. 2003;Kress et al. 2015). The plethora of studies now using solely DNA- based species identification to link life history stages and to study trophic interactions is a convincing testimony (see references in BIntroduction^). In both cases, the sampling effort is comparable, however, with slightly different approach. While the collecting of active postlarval forms for the purpose of molecular delimitation does not need to focus on ovigerous females, the arthropod-parasitizing larvae require complementary collecting methods.
Although rearing gives a vital insight in the species biology, the disadvantage is the time. The faster genetic matching has its flaws in laboratory costs and requires access to molecular facilities. On the other hand, the per base pair sequencing cost has plummeted in the last 10 years, thanks to highthroughput Sanger sequencing workflows as well as nextgeneration sequencing techniques. DNA-based species identification also has a number of challenges in the need of a representative reference library and can be misled by hybridization, introgression, incomplete lineage sorting, or young species radiation (see BIntroduction^). Species delimitation using the GMYC model can be sensitive to low sampling, both at population level and at taxonomic scale (Lohse 2009). In our case, the inclusion of additional species of the same genus resulted in increased support in the gene tree and gave better performance in the species delimitation (narrowing the confidence interval). Also the possibility of contamination (gut content from host/prey) is harder to detect with fewer samples. As we observed, conspecific parasitism under laboratory conditions, it is not impossible that parasitism can occur in the wild between closely related species. DNA contamination from gut contents in such case could be very difficult to detect. Of several molecular species delimitation models, the GMYC model has the advantage of not requiring any predefined units. It is a so-called discovery method of species delimitation in contrast to validation methods that require a predefined hypothesis of species limits (Ence and Carstens 2011;Carstens et al. 2013). In this study, the combination of experimental rearing and molecular matching contributed to the knowledge of intraspecific variation of metric data and strengthened the conclusions resulting in formal synonymization of several nominal species names. This is a first step on the way towards resolving the double classification in Erythraeus and other parasitengone mites. As stated by de Queiroz (2007), assembling morphological, biological, and molecular data, gives multiple lines of evidence to species hypotheses in a unified species concept framework.