Phylogenetic relationships and subgeneric classification of European Ephedrus species (Hymenoptera, Braconidae, Aphidiinae)

Abstract In this study two molecular markers were used to establish taxonomic status and phylogenetic relationships of Ephedrus subgenera and species distributed in Europe. Fifteen of the nineteen currently known species have been analysed, representing three subgenera: Breviephedrus Gärdenfors, 1986, Lysephedrus Starý, 1958 and Ephedrus Haliday, 1833. The results of analysis of COI and EF1α molecular markers and morphological studies did not support this classification. Three clades separated by the highest genetic distances reported for the subfamily Aphidiinae on intrageneric level. Ephedrus brevis is separated from persicae and plagiator species groups with genetic distances of 19.6 % and 16.3 % respectively, while the distance between persicae and plagiator groups was 20.7 %. These results lead to the conclusion that the traditional subgeneric classification of Ephedrus needs revision. Species from persicae species group are raised to subgenus level as Fovephedrus Chen, 1986 and Lysephedrussyn. nov. is assigned as a junior synonym of subgenus Ephedrus. Key for identification of Ephedrus subgenera is provided. Ephedrus hyadaphidis Kocić & Tomanović sp. nov. is described and several species are confirmed as valid species for the first time. Furthermore, two species are synonymised: E. dysaphidissyn. nov. as a junior synonym of E. cerasicola and E. blattnyisyn. nov. as a junior synonym of E. plagiator.


Introduction
Members of the subfamily Aphidiinae (Hymenoptera, Braconidae) display a fascinating life cycle as obligatory koinobiont parasitoids of aphids, regulating host population size and density and therefore they are considered as a beneficial insect group. Due to their importance in biological control of aphids, this is one of the most extensively investigated groups within the family Braconidae (Quicke 2015), yet their phylogeny and taxonomy still remains unresolved.
Several species from this genus were implemented in biological control programs, and some of them are commercially produced and packed as part of the cocktails containing different parasitoid agents (Menten 2011). Two of Ephedrus species are commercially sold in great numbers, E. cerasicola (hundreds of thousands to one million of specimens per week) and somewhat less E. plagiator (ten thousand to hundred thousand per week) (van Lenteren et al. 2018). Moreover, there have been attempts of Ephedrus species introductions to different continents in order to control alien aphid species with the same geographical origin (Autrique et al. 1989, Carver 1989, Starý 1993, Elliott et al. 1995, Ripa et al. 1995, Armstrong and Peairs 1996. Unfortunately, in the majority of these introductions, Ephedrus populations were not established. Systematic position of Ephedrus within the subfamily Aphidiinae is still uncertain. There have been numerous hypotheses about which genus or group is most related to Ephedrus. Based on wing venation and several other symplesiomorphic characters, the genera Toxares Haliday, 1833 and Ephedrus were classified within the tribe Ephedrini (Mackauer, 1968), yet studies on larval morphology showed that they differ significantly (Finlayson 1990). Furthermore, morphological examination of the first instar larvae indicated great morphological similarity between Ephedrus and Praon Haliday, 1833(O'Donnell 1989. The genus Parephedrus Starý & Carver, 1971 from Australia is considered related to Ephedrus (Starý and Carver 1971;Mackauer and Finlayson 2012), sharing similar wing venation and 11 segmented antennae in both sexes. Although Schlinger (1974) positioned Pseudephedrus Starý, 1972 and Parephedrus within the subfamily Ephedrinae (now tribe Ephedrini), these two archaic genera together with Vanhartenia van Harten, 1974 andChoreopraon Mackauer, 2012 are considered to form a separate branch in the phylogeny of Aphidiinae (Starý 1976).
The second uncertainty about the systematic position of the genus Ephedrus is whether or not it represents the basal group within the subfamily Aphidiinae. There are some, mostly molecular, studies which consider the tribe Praini as basal within the subfamily (Smith et al. 1999), while some others listed Aclitus (Kambhampati et al. 2000) or even Pseudephedrus (Žikić et al. 2017b) as basal genera. On the other hand, numerous molecular and morphological studies suggest that Ephedrus is the most likely candidate for taking a basal position (Gärdenfors 1986, Belshaw and Quicke 1997, Sanchis et al. 2001, Shi and Chen 2005. Species of this genus possess a range of plesiomorphic characters, such as forewing with fully developed braconid venation, 11-segmented antennae in both sexes, existence of central areola of propodeum, shape of ovipositor sheaths, short petiole and several more (Gärdenfors 1986). Additional evidence that goes into favour of its basal position is the discovery of fossil specimens resembling species of this genus, namely Ephedrus mirabilis Timon-David, 1944and Ephedrus primordialis Brues, 1933(Starý 1973, Ortega-Blanco et al. 2009).
Until now, species of Ephedrus have only been used for molecular studies on the subfamily level (Belshaw and Quicke 1997, Smith et al. 1999, Kambhampati et al. 2000, Sanchis et al. 2001, Žikić et al. 2017b, while taxonomic status of its subgenera and relationships between species have never been tested with molecular markers. Currently, genus Ephedrus comprises three subgenera: Lysephedrus (E. validus), Breviephedrus (E. brevis) and Ephedrus (all other species), and the third is divided into three species groups (plagiator, lacertosus, persicae) (Gärdenfors 1986). Based on the presence of fovea on mesoscutum, Chen (1986) described the genus Fovephedrus Chen, which was later synonymised with Ephedrus, but proposed as a separate subgenus (van Achterberg 1997). However, position of Fovephedrus within genus Ephedrus was completely unclear.
We decided to investigate the taxonomic status and phylogenetic relationships of Ephedrus subgenera and species with European origin, with the combination of nuclear and mitochondrial markers and morphology. Here we propose a new subgeneric classification. Several species in this study are confirmed by molecular approach for the first time and phylogenetic relationships among European species of genus Ephedrus are presented. We describe Ephedrus hyadaphidis sp. nov., an additional member of the plagiator group, parasitoid of Hyadaphis foeniculi Passerini, 1860 on several plant species. Furthermore, we synonymise E. dysaphidis as a junior synonym of E. cerasicola and E. blattnyi as a junior synonym of E. plagiator.

Sample collection and morphological analysis
Parasitoid specimens were sampled throughout Europe during the past two decades. Sampling was conducted in two ways, by net sweeping and by rearing of the parasitoids. The second method was preferred as rearing provides important data on their tri-trophic interactions. Parts of the plants infested with aphid colonies were stored in plastic containers with the lid openings covered with mesh. The samples were transported to laboratory and kept under controlled conditions until the emergence of the parasitoids. Aphids stored in 70% ethanol and plant samples were also collected and identified to species or genus level. Unfortunately, all specimens of E. lonicerae were slide mounted and thus couldn't be used for molecular analyses (Žikić et al. 2009) and after a thorough search, we were not able to locate and acquire specimens of E. vaccinii or E. longistigmus.
Each specimen was examined under the ZEISS Discovery V8 stereomicroscope (Carl Zeiss MicroImaging GmbH, Gottingen, Germany). After DNA extractions, samples were dissected and slide-mounted in Berlese medium. When available, specimens were studied using the Jeol JSM-6460LV scanning electron microscope (Jeol USA, Inc., Peabody, MA, USA). Measurements for new species description were obtained using ImageJ software (Schneider et al. 2012) based on photographs taken with Leica DM LS phase contrast microscope (Leica Microsystems GmbH, Wetzlar, Germany). Morphological terminology follows Sharkey and Wharton (1997).
The examined material is deposited in the Insitute of Zoology, Faculty of Biology, University of Belgrade (Serbia), except for specimens of E. validus and E. koponeni that were loaned from the Zoological Museum, University of Helsinki (Finland) and 5♀9♂ of E. hyadaphidis paratypes that are deposited in the Croatian Natural History Museum, Zagreb, Croatia. Ephedrus specimens analysed in this study are presented in Suppl. material 1.

Molecular analysis
Total genomic DNA was extracted from single individuals using the Qiagen Dneasy Blood & Tissue Kit (Qiagen, Valencia, CA, USA), following the manufacturers' protocol. The extraction method was non destructive in order to preserve whole specimens that could be used in further studies of external morphology. Two molecular markers were used; mitochondrial cytochrome c oxidase subunit I (COI) and nuclear elongation factor 1 alpha (EF1α). The primers used for amplification of COI and EF1α fragments were LCO1490 and HCO2198 (Folmer et al. 1994) and EFf and EFr (Belshaw and Quicke 1997), respectively. The amplification was carried out in a total volume of 40 µl which contained 2 µl of extracted DNA, 0.4 µl of Taq polymerase, 2 µl of each primer (10 mM), 2.4 µl of dNTPs (0.6 mM), 3.6 µl of MgCl 2 , 4 µl of buffer and 23.6 µl of nuclease free water. PCR profile for COI was: 5 minutes of initial denaturation followed by 35 cycles of 60 second denaturation (94 °C), 60 second annealing (54 °C) and 90 second extension (72 °C) and 7 minutes of final extension (72 °C). PCR conditions for amplification of EF1α were same as in Belshaw and Quicke (1997). Obtained PCR products were purified with QIAquick PCR Purification Kit (Qiagen, Valencia, CA, USA) and sent to Macrogen Inc. (Seoul, Korea) for sequencing. DNA of dry museum specimens, due to fragmentation of DNA, was amplified by using primers for amplification of short barcoding fragments, following the protocol by Mitrović and Tomanović (2018). For certain specimens the quantity of extracted DNA was too low, so only the COI fragment was obtained. That was the case with E. brevis individuals. For others, like E. chaitophori and E. tamaricis, the amplification of EF1 alpha fragment yielded PCR products of high quality, but the sequencing was unsuccessful. The reason for this might be the interruption of sequencing due to the existence of DNA introns or a deletion along the fragment.

Phylogenetic analysis
The sequences acquired after amplification were checked for pseudogenes, visualised in FinchTV Geospiza Inc. (Seattle, USA) and manually edited and aligned in BioEdit program (Hall 1999). Sequences were trimmed to the same length of 557 and 440 base pair positions for COI and EF1α, respectively. MEGA 6 (Tamura et al. 2013) software was used to estimate the evolutionary divergence between sequences. The analysis for both gene fragments was conducted independently in BEAST 2.5 (Drummond et al. 2012) software platform for Bayesian evolutionary analysis. The initial data file was designed in BEAUti with the Tamura-Nei (Tamura and Nei 1993) model with gamma distributed rates among sites (TN93+G), which MEGA 6 proposed as the most appropriate model. The initial data file conducted the analysis with strict clock type and Yule Process speciation. The species Venturia canescens Gravenhorst, 1829, belonging to the sister family Ichneumonidae was used as an outgroup species in all analyses. Obtained phylogenetic trees which contained all sequences visualised by FigTree 1.4.3. software (Rambaut 2009) were large and hard to follow, thus a single sequence of each haplotype was used to present the results.

Results
Phylogenetic analyses of the two molecular markers yielded trees with similar branch topographies (Figs 1, 2). A total of 86 COI sequences was analysed (seven, including outgroup, were mined from GenBank and 18 were used from authors' previous publications), and 58 different haplotypes were detected (Fig. 1). Species E. brevis, E. cerasicola, E. laevicollis, E. validus, and E. prociphili are confirmed as distinct species for the first time with the barcoding marker. Phylogenetic tree based on COI sequences showed species separation into three main phylogenetic clades, which does not correspond to traditional subgenera delineation (Fig. 1). First clade was formed by species belonging to persicae species group (E. persicae, E. chaitophori, and E. tamaricis). Second clade was represented with E. brevis and third with all other species (plagiator clade). Ephedrus brevis, the only representative of Breviephedrus subgenus was separated from both persicae and plagiator clades on the phylogenetic tree with the average genetic distances of 19.6 % and 16.3 %, respectively, while genetic distance between persicae clade and plagiator clade was 20.7 % (Suppl. material 2).
According to the obtained results, the nominative subgenus Ephedrus is paraphyletic, consisting of two independent lineages, plagiator and lacertosus species groups. In addition to these groups E. validus, the only member of the Lysephedrus subgenus, also grouped within this clade. Moreover, E. validus is nested within the plagiator species group, forming a separate clade together with E. helleni. The second clade within the plagiator species group consists of specimens belonging to ten different species (Ephedrus blattnyi, E. cerasicola, E. dysaphidis, E. hyadaphidis sp. nov., E. koponeni, E. laevicollis, E. nacheri, E. niger, E. plagiator, and E. prociphili). One individual determined as E. blattnyi grouped with E. plagiator specimens, with the genetic distance ranging from 0.2 % to 0.7 %. The same was the case for one E. dysaphidis specimen which clustered within the E. cerasicola clade (genetic distance 0.0 %-1.6 %). Genetic distances between other species within this group vary greatly ranging from 1.1 % between E. nacheri and E. prociphili up to 7.6 % between the previous two and E. laevicollis (Table 3). Ephedrus lacertosus is clearly separated as a distinct group with genetic distances from all other Ephedrus species above 8.9 %. The clade consisting of Ephedrus persicae species is the oldest within the genus and represents a separate subgenus. Within this clade all three analysed species were confirmed as valid. Additionally, within E. persicae two separate phylogenetic lines were determined with average genetic distance of 2.5 % (Fig. 1).
With the second molecular marker EF1α, 14 sequences (four mined from Gen-Bank), belonging to eleven species (Ephedrus cerasicola, E. helleni, E. hyadaphidis sp. nov., E. lacertosus, E. laevicollis, E. nacheri, E. niger, E. persicae, E. plagiator, E. prociphili, E. validus) were analysed. While the genetic distances between species groups and species were significantly lower than those based on COI sequences, the topography of the phylogenetic three remained similar (Fig. 2). As previously, sequences separated into two groups, the first composed solely out of the representative of persicae species group, E. persicae, and the second containing species from lacertosus and plagiator groups. A significant change in relation to the COI phylogenetic tree is clustering of E. validus and E. helleni together with E. lacertosus, species from lacertosus group. This topography could be explained by genetic distances: E. helleni and E. validus show genetic distances of 3.8-7.2 % and 5.7-9.0 %, respectively, when compared to other plagiator species, whereas compared to lacertosus, these distances are 4.2 % and 5.4 %, respectively. All genetic distances were considerably lower compared to those based on COI, since EF1α is more conservative nuclear molecular marker (see Suppl. material 2). The average genetic distances between E. persicae and lacertosus and plagiator species groups were 7.6 % and 9.0 %, respectively.

Description of the new species
Etymology. Name of the species is derived from its aphid host, Hyadaphis foeniculi. Distribution. The current species distribution is Balkan Peninsula. Depositories. Holotype is slide mounted and deposited in the collection of the Institute of Zoology, Faculty of Biology, University of Belgrade. Paratypes collected in National Park Plitvice, Croatia are deposited in the Croatian Natural History Museum, Zagreb, Croatia. The remaining paratypes are deposited at the Institute of Zoology, Faculty of Biology, University of Belgrade.
Synonymies. The molecular phylogenetic analysis clustered E. dysaphidis together with E. cerasicola, with the genetic distance ranging from 0.0% to 1.6%. We performed a detailed morphological examination of all available material of both species in order to test the obtained molecular results. Ephedrus dysaphidis is a species from plagiator species group, described in the study by Tomić et al. (2005). Authors differentiated it from other species in this group by number of longitudinal placodes on F1 (1-2, rarely 3) and F2 (2-3, rarely 4) and shorter petiole (1.94-2.20 times as long as wide). The colour of scapus, pedicel and the ring at the base of F1 is stated as brownish to yellow. It is a parasitoid of aphids from genus Dysaphis Börneron on Malus domestica Borkh. While studying morphology of many additional populations of E. dysaphidis and E. cerasicola, we found numerous overlapping characters. In the description of E. cerasicola, Starý (1962) states that F1 is more than three times as long as wide, while in the paper of Tomić et al. (2005) the length/width ratio of E. dysaphidis is 3.8-4.6. Furthermore, F1/F2 length ratio is also overlapping in both species (in E. cerasicola F1 is longer by 1/3 than F2 and in E. dysaphidis that ratio is 1.28). Propodeum in both species is with a pentagonal areola, having the same number of setae on upper and lower areolae. Additionally, pterostigma in E. cerasicola is more than four times as long as wide; in E. dysaphidis this ratio varies between 4.1-4.7. Finally, petiole in E. cerasicola is less than twice as long as wide, while in E. dysaphidis, according to the authors, this ratio is 1.94-2.20. However, our studied material of both species showed that the length to width ratio at the spiracle level of E. cerasicola and E. dysaphidis is 1.97-2.1 and 1.90-2.18, respectively, thus also overlapping. The only character that differs in two descriptions of species is the number of longitudinal placodes on F1 and F2 flagellar segments, which is 0 and 2 in E. cerasicola and, as mentioned above, 1-2(3) and 2-3(4) in E. dysaphidis, respectively. However, we found specimens from the same sample (that were first identified as E. dysaphidis based on the aphid host Dysaphis plantaginea Passerini, 1860 on Malus domestica) having a different number of longitudinal placodes that varied from F1:0, F2:1 to F1:1, F2:3. Moreover, in the revision of the genus Ephedrus Gärdenfors (1986) states that the number of longitudinal placodes in F1 and F2 for E. cerasicola is 1 and 2-3, respectively. One more distinction between the species was the colour of scape, pedicel and F1 and F2. In E. cerasicola scape, pedicel, F1 and part of F2 are yellow brownish (Starý 1962) or can be with scape yellowish to brown, pedicel and at least base or the entire F1 yellow, while F2 is yellowish at base (Gärdenfors 1986). While examining the material of both species we found various gradations of scape, pedicel, F1 and F2 colour, mainly following the description of Starý (1962) and Gärdenfors (1986 Ephedrus blattnyi is a specialised parasitoid described from one finding in the Czech Republic, reared from Pterocomma ringadhli (junior synonym of Pterocomma rufipes Hartig, 1841) on Salix caprea. Several authors during previous years questioned the validity of E. blattnyi species status. After the examination of type specimens Gärdenfors (1986) stated that they are extraordinarily similar to E. plagiator and that it is possible that they represent aberrant specimens of E. plagiator due to an unusual aphid host. Furthermore, Koponen and Halme (1993) reported that they collected specimens fully corresponding to the E. blattnyi redescription (Gärdenfors, 1986). However, the authors didn`t include them in the paper, since the distinguishing characters were not reliable. Finally, Tomanović (2000) states, while reporting the finding of specimens corresponding to E. blattnyi, that it is very similar to E. plagiator and E. prociphili. We analysed specimens from P. rufipes which morphologically corresponded to the description of E. blattnyi reared from P. rufipes aphid host on Salix retusa. Molecular analysis of COI clustered E. blattnyi within the E. plagiator, with the molecular distance ranging from 0.2% to 0.7%. Therefore, we conclude that E. blattnyi represents a morphotype of E. plagiator and assign it a status of junior synonym.

Subgenera and species groups of the genus Ephedrus in Europe
Analysing all obtained results (both molecular and morphological), we concluded that the current subgeneric classification of Ephedrus needs revision and here we propose a new one. Subgenus Lysephedrus Starý 1958 is synonymised as a junior synonym of the subgenus Ephedrus Haliday, 1833 from which persicae species group is raised to the level of the subgenus Fovephedrus Chen, 1986. The subgenera Fovephedrus and Ephedrus are redescribed.

Discussion
Continuously advancing molecular methods allow us to easily sequence gene fragments of interest and compare them with results of morphological and ecological analyses in order to obtain a clearer picture of phylogenetic relationships and evolution of the group of interest. In this study we used two molecular markers to reveal the taxonomic positions and phylogeny of species within the genus Ephedrus in Europe. As proposed by Derocles et al. (2012), we used a combination of both mitochondrial and nuclear markers. The barcoding region of mitochondrial cytochrome c oxidase subunit I proved to be a reliable marker in resolving taxonomic relationships in previous studies of the subfamily Aphidiinae (Folmer et al 1994, Petrović et al. 2016, Mitrović et al. 2019). Furthermore, we conducted an analysis of nuclear EF1α that mostly supported the inter-specific relationships of our taxa previously provided by COI. Previous studies classified species from genus Ephedrus into three subgenera (Ephedrus, Breviephedrus (E. brevis) and Lysephedrus (E. validus)) based on morphology (Gärdenfors 1986). However, with the combination of molecular and morphologi-cal approach we here revise the subgeneric classification. All three revised subgenera separated with the highest genetic distances known in Aphidiinae when analysing subgenera, ranging from 16.3 % to 20.7 %. Currently, the only member of the subgenus Breviephedrus, E. brevis, is at first sight easily distinguished from any other Ephedrus species with a very thickset, stocky and black polished body (Stelfox 1941). It is interesting to note that the parasitoid has never been reared from an aphid host (which is still unknown), but is suspected to be from the aphid species associated with Betula, since the specimens are found by sweeping or by traps in the vicinity of birch trees. Gärdenfors (1986) states this species as the most primitive within the genus, morphologically similar to Parephedrus. Indeed, E. brevis possesses a wide range of morphological characters that are also found in Parephedrus, thus indicating that these two could be a "connection bridge" in the evolution of ancient genera Parephedrus, Vanhartenia, Pseudephedrus, and Choreopraon on one side and the rest of Aphidiinae on the other. In order to uncover phylogenetic relationships between these groups, future studies are needed, especially in the light of our molecular results which do not support previous statements that it is the most primitive species within the genus.
Based on the presence of the fovea on mesoscutum, Chen (1986) described the genus Fovephedrus, with the type species Fovephedrus radiatus Chen, 1986. Several E. persicae morphotypes (E. persicae, E. palaestinensis (= E. persicae, see Gärdenfors (1986)), E. rugosus, E. radiatus, E. transversus) and E. longistigmus (member of lacertosus species group) were later reclassified to this genus, based on the same morphological character (Chen and Shi 2001). Van Achterberg (1997), synonymised Fovephedrus as a synonym of Ephedrus, but stated that it could be considered as a valid subgenus. The presence of mesoscutellar fovea is characteristic of several Ephedrus species. Even within E. persicae its presence is very variable: fovea can be absent, or the specimens can possess one (the most common state) or two foveal pits. Furthermore, this morphological character is shared with the lacertosus species group, where it can sometimes be found in E. lacertosus specimens, and is always present in E. longistigmus. In the end, the existence of the mesoscuteal fovea is not a character specific for species of genus Ephedrus, but is also found among Toxares (Tomanović et al. 2008). We consider this morphological character to be diagnostically unreliable, especially at the genus level.
The redescribed subgenus Fovephedrus is raised from persicae species group and its inter-specific relationships are thoroughly discussed in authors' recent study (Petrović et al. 2016). We did not examine type species E. radiatus but judging by the description and drawings (Shi and Chen 2001) it obviously belongs to Ephedrus persicae species group. The same is the case with Asiatic species E. rugosus and E. transversus which on the basis of descriptions both belong to persicae group. As mentioned earlier, E. palaestinensis is synonymised with E. persicae (Gärdenfors 1986) and E. longistigmus apparently does not belong to subgenus Fovephedrus, but lacertosus group within the traditional subgenus Ephedrus. The subgenus Ephedrus holds the remainder of the species and is divided into two species groups, plagiator and lacertosus. While E. longistigmus is reported only two times (see Koponen andHalme 1993, Davidian 2018), E. lacertosus is common representative of the lacertosus group in Europe. This species group is considered to be the most phylogenetically advanced within the genus Ephedrus (Gärden-fors 1986), due to several apomorphic morphological traits, like an elongated petiole, flagellomere 1 and pterostigma. Our results confirm the separation of E. lacertosus into a distinct clade. Within the plagiator group, E. helleni and E. validus position as most distant phylogenetically from all other plagiator species. Both parasitoids are specialists; first one attacking Cavariella aphids across its distribution (Koponen and Halme 1993, Kavallieratos et al. 2004, Rakhshani et al. 2007, Starý and Havelka 2008) and additionally Eumyzus Shinji, 1929 in Asia (Davidian 2007, Davidian andGavrilyuk 2014) and the second one parasitising root aphids from the subfamily Eriosomatinae (Gärdenfors 1986). Starý (1958) raised E. validus to the subgenus Lysephedrus, based on morphological characters, such as very rugose propodeum and petiole, shape of ovipositor sheaths and their heavy pubescence. Davidian considered it as separate genus (Davidian 2007). However, since this species is a parasitoid of waxy root aphids, all those differences could represent adaptations to the specific underground ecological niche (Gärdenfors 1986). It is probable that E. validus and E. helleni separated from other species early in the evolution of this group, driven by a specialisation for a certain aphid host. The results of molecular analysis of nuclear EF1α grouped E. lacertosus, E. helleni, and E. validus into one clade, separating them from the rest of plagiator species.
The fact that revealing new cryptic species within large generalist groups is an ongoing process (Derocles 2016, Petrović et al. 2016) is once again proven by the description of the additional member of plagiator group, E. hyadaphis. The specimens reared from Hyadaphis foeniculi were collected across the Balkans during the period of fifteen years, from different plant hosts (Lonicera xylosteum, Sanicula europaea, and Anthriscus sylvestris), all common for this aphid species. Ephedrus hyadaphidis parasitoids were found in the woodland area on higher altitudes (524-1140 meters above sea level). This species might have been noticed earlier: in his revision Gärdenfors (1986) discusses an Ephedrus specimen collected in Italy, reared from Hyadaphis foeniculi on Lonicera etrusca, similar to E. nacheri, but with several morphological differences. Along with E. lonicerae and E. tamaricis, E. hyadaphis represents the third recently described species of Ephedrus distributed in the Balkan region. Furthermore, E. persicae is separated into two clades, one containing Mediterranean taxa, and the other taxa from the rest of Europe. The average genetic distance between these two clades is 2.5 %, suggesting that they have been evolving separately for some time. Thus, both clades require further taxonomic investigation. The richness of species that are distributed only in the Balkan region, as well as richness of haplotype numbers might imply that this area played a significant role for Ephedrus species as a refuge during the glacial periods.
Ephedrus koponeni is reported outside its known distribution, northern Europe (Finland and European part of Russia) (Halme 1992, Davidian 2018. One male and one female were reared from Cinara sp. on the Balkan pine, Pinus peuce, a relict and endemic species with a fragmented distribution in the Balkan Peninsula (in Serbia inhabiting only very southern part). It is possible that the specimens of this species across Europe are, when seldom collected, misidentified as E. plagiator. The other probability is that E. koponeni possesses an extremely fragmented distribution, restricted only to the secluded areas of its former species range.
The results of this study show close phylogenetic relationships among E. koponeni, E. prociphili, and E. nacheri. All three species are morphologically very similar to E. plagiator, distinguished from each other by subtle morphological differences and aphid host range (Gärdenfors 1986). While E. koponeni and E. prociphili are specialised to coniferous aphid species and Prociphilus sp., respectively, E. nacheri has a somewhat broader aphid host range, parasitising several species from Aphidini and Macrosiphini tribes. In Europe this species is most frequently reared from the genus Hayhurstia Del Guercio, 1917. These species might be quite young in the evolution line of the genus Ephedrus, just recently separated from the others in plagiator species group, which would explain lower genetic distances and great morphological similarity with E. plagiator. Furthermore, our study suggests that E. cerasicola (parasitoid of Myzus and Dysaphis) and E. laevicollis (parasitoid of Chaetosiphon Mordvilko, 1914), are closely related species.
In summary, our results show that the phylogeny of Ephedrus is more complex than previously thought. It is important to note that European species of Ephedrus comprise less than a half of currently described species, so in order to get complete insight into the phylogenetic relationships among species and their evolution, further studies are needed.
Petr Starý (Laboratory of Aphidology, Institute of Entomology, Academy of Sciences of the Czech Republic) for E. prociphili samples. We thank Katica Poljak for remarkable help in collecting material in NP Plitvice.