Review of the sawfly genus Empria (Hymenoptera, Tenthredinidae) in Japan

Abstract The following eleven Empria species are reported from Japan: Empria candidata (Fallén, 1808), Empria japonica Heidemaa & Prous, 2011, Empria liturata (Gmelin, 1790), Empria loktini Ermolenko, 1971, Empria plana (Jakowlew, 1891), Empria quadrimaculata Takeuchi, 1952, Empria rubicola Ermolenko, 1971, Empria tridens (Konow, 1896), Empria tridentis Lee & Ryu, 1996, Empria honshuana Prous & Heidemaa, sp. n., and Empria takeuchii Prous & Heidemaa, sp. n. The lectotypes of Poecilosoma pallipes Matsumura, 1912, Empria itelmena Malaise, 1931, Tenthredo candidata Fallén, 1808, and Tenthredo (Poecilostoma) hybrida Erichson, 1851 are designated. Empria itelmena Malaise, 1931, syn. n. is synonymized with Empria plana (Jakowlew, 1891). Poecilosoma pallipes Matsumura, 1912, previously assigned to Empria, is transferred to Monsoma, creating Monsoma pallipes (Matsumura, 1912), comb. n. Results of phylogenetic analyses using mitochondrial (COI) and nuclear (ITS1 and ITS2) sequences are also provided.

introduction With 51 valid species-level taxa (Taeger et al. 2010;Prous et al. 2011b), Empria Lepeletier & Serville, in Latreille et al. 1828 is one of the largest genera in the Allantinae. Nevertheless, it still remains rather poorly studied in comparison with other tenthredinid sawflies. Empria species are often misidentified because of the lack of easily observable diagnostic characters. Fortunately, their genitalia frequently possess clear differences even between closely related species mostly enabling their reliable identification. Though the knowledge on most of the European Empria species can be regarded as satisfactory (Zhelochovtsev and Zinovjev 1988;Prous et al. 2011b), very little is known about Eastern Palaearctic species. According to Takeuchi (1952a), more than seven Empria species had been found in Japan, but most of them remained unidentified. Until recently, only two species had been identified (Takeuchi 1952a;b;Abe and Togashi 1989), and one of them, Empria pallipes (Matsumura 1912), actually belongs to Monsoma MacGillivray, 1908 (see results). Prous et al. (2011b) reported three additional species. Here we report 11 species from Japan, two of them described as new. One male, probably representing a new Empria species (sp. 1) is also discussed but not yet described as new due to insufficient material.
No attempts to reconstruct the phylogeny of Empria have been made so far. Some preliminary results based on a limited number of species can be found in Prous et al. (2011b), which focuses on the E. longicornis species group. Only few intrageneric groups have been proposed, which might be monophyletic. In particular, Empria is sometimes divided into the subgenera Parataxonus MacGillivray, 1908 [now comprising E. candidata (Fallén, 1808) and E. multicolour (Norton, 1862)] and Empria s. str. (all other species) (Ross 1936;Zhelochovtsev and Zinovjev 1988;Yan et al. 2009). Within Empria s. str., the E. hungarica (Konow, 1895) (see Heidemaa and Viitasaari 1999) and the E. longicornis (Thomson, 1871) species groups (see Prous et al. 2011b) have been proposed. In addition, the E. immersa species group can be defined for the species possessing highly similar penis valves, which have a characteristic long apical spine (Smith 1979;Zhelochovtsev and Zinovjev 1988;Prous et al. 2011b). To examine the phylogenetic relationships within Empria based on DNA sequences, we here expand the dataset of Prous et al. (2011b) by including 7 more species (six outside and one inside of the longicornis-group). For this, we use one continuous mitochondrial region (full COI, two complete, and one incomplete tRNAs) and one nuclear region (ITS1 and ITS2 within the rRNA locus) analysed separately and in combination using Bayesian methods.

Material and methods
Pinned specimens studied are from the following institutional collections: BMNH Natural History Museum, London, United Kingdom (G. Broad, N. Dale-Skey Papilloud, S. Ryder, N. Springate); Specimens from the private collections of Erik Heibo, Guy T. Knight, and of the second author (MH) were also studied.
For morphological analyses, male penis valves, female lancets (valvula 1), and external characters of the adults were studied.
To dissect the penis valves, genital capsules were separated from the specimen and macerated in KOH or NaOH (10-15%) for 6-12 hours at room temperature, or treated with proteinase K using High Pure PCR Template Preparation Kit (Roche, Mannheim) and following manufacturer's protocol.
Imaging methods are described in Prous et al. (2011b). All images made for this study are deposited in the Morphbank database (http://www.morphbank. net/?id=592670).
For molecular phylogenetic analyses, DNA sequences of the internal transcribed spacers 1 and 2 (ITS1 and ITS2), and a mitochondrial DNA (mtDNA) fragment containing tRNA-Cys, tRNA-Tyr, cytochrome c oxidase I (COI), and partial tRNA-Leu, were obtained using methods described in Prous et al. (2011b). However, because amplification of ITS2 of Empria honshuana sp. n. failed using the primers CAS5p8sFc and CAS28sB1d (Ji, Zhong and He 2003;Prous et al. 2011b), we used the primers AM1 (5´ TGT GAA CTG CAG GAC ACA TGA 3´) and AM2 (5´ATG CTT AAA TTT AGG GGG TAG TC 3´) (Marinucci et al. 1999;Heidemaa et al. 2004) instead. The PCR programme in this case consisted of an initial denaturing step at 95°C for 1 min, followed by 43 cycles of 20 s at 95°C, 30 s at 65-55°C (a touchdown profile was used, in which the annealing temperature decreased from 65°C to 55°C by 0.5°C every cycle), and 70 s at 68°C; the last cycle was followed by a final 7 min extension step at 68°C. For some older air-dried museum specimens, it was possible to obtain the sequences only partially. Sequences reported here have been deposited in the GenBank (NCBI) database (accession numbers JN029842-JN029898). As suggested by Chakrabarty (2010), DNA sequences from type material are here referred to as genetypes.
Boundaries of the sequenced tRNA and ITS2 genes were identified as described by Prous et al. (2011b). Phylogenetic analyses of ITS genes were performed using Bali-Phy 2.0.2 (Suchard and Redelings 2006) since this program has implementations to handle difficult-to-align sequences. In order to enhance the speed of calculation, sequences were aligned manually for detecting and fixing the conserved positions prior to analysis with Bali-Phy. Four independent analyses were run (203 213-262 061 iterations) using the GTR + I + G[4] model. The first 10 000-60 000 iterations were discarded as "burn in" after examination of log-likelihood scores in Tracer 1.4 (available from http://beast.bio.ed.ac.uk/Tracer).
Phylogenetic analysis of the mitochondrial genes and combined analysis of the nuclear and mitochondrial genes were performed with MrBayes 3.1.2 (Huelsenbeck and Ronquist 2001;Ronquist and Huelsenbeck 2003) using the GTR + I + G[4] model. Mitochondrial sequences were aligned manually, and prior to phylogenetic analyses, non-coding and ambiguously aligned tRNA regions, one insertion of three base pairs in COI of Monsoma pulveratum (Retzius, 1783), and two to three amino acid coding codons of COI at the 3´ end (the last three codons of E. quadrimaculata and E. rubicola could not be unambiguously aligned with the last two codons of other species) were excluded. In the combined analysis we used MAP (maximum a posteriori) alignment of ITS obtained from one of the four analyses with Bali-Phy. Both mitochondrial and combined datasets were run for 5 000 000 MCMC generations, with trees and lnL's sampled at intervals of 100 generations. The first 25% of generations were discarded as "burn-in". Monsoma pulveratum was used to root the trees.

Data resources
The data underpinning the analyses reported in this paper are deposited in the Dryad Data Repository at doi: 10.5061/dryad.fs262s48 (Prous et al. 2011a) and at GBIF, the Global Biodiversity Information Facility, http://ipt.pensoft.net/ipt/resource. do?r=japanese_empria.
Notes. Male unknown. Matsumura (1912) did not give the number of specimens he used for the original description. A female syntype bearing a red type label is hereby designated as the lectotype. Taxonomic affinities. The morphologically closest species is the Nearctic E. multicolor, from which E. candidata can be distinguished by the following characters: femora predominantly and most other parts of legs at least partly black (legs are almost entirely yellowish in E. multicolor), tarsal claws simple or with a minute inner tooth (with a long subbasal tooth in E. multicolor), shallowly emarginated clypeus (deeply emarginated in E. multicolor), and postocellar area more than 1.6 times wider than long (less than 1.5 in E. multicolor) (see also Smith 1979 Prous et al. 2011b).
Head. Head behind eyes in dorsal view subparallel sided; postocellar area trapeziform, its length equal to or longer than 2 times diameter of lateral ocellus; distinct and diverging lateral postocellar furrows going from ocelli towards occiput at least to the distance of ocellus diameter; area between frontal crests clearly exceeding the level of crests in dorsal view; postocellar area with indistinct punctures and interspaces, more or less glossy; punctures more regular on temples and postocular area, face with more irregular punctures; wrinkled interspaces more prominent on frontal area; clypeus with rough irregular punctures, more or less fused; ocellar and postocellar area convex, slightly raised; clypeus tridentate with median keel distinct mostly in anterior part of clypeus only, median tooth smaller than lateral teeth; malar space about equal to or shorter than distance between antennal sockets; frontal ridge V-shaped; pit in central part of frontal field present; median ocellus surrounded by groove, with short distinct longitudinal furrow anteriorly, and with similar but mostly less distinct furrow posteriorly. Maximal length of temple 1.2-1.4 times greater than its minimal length; flagellum 1.9-2.0 times longer than breadth of head.
Thorax. Mesoscutellum, mesoscutellar appendage, and metapostnotum more or less glossy, almost impuctate or with indistinct shallow punctures; metascutellum with irregular fine punctures; punctures on mesoscutum more evident on lateral and anterior regions of the median lobes, fading towards central regions; mesepisternal punctures variable between specimens, from rather weak with intespaces almost glossy to more distinct with sculptured, interspaces; mesepimeron with setae on posterior part; metepisternum with evenly distributed setae; metepimeron in central part without setae; distance between cenchri 1.1-1.4 times of cenchrus width; wings hyaline, venation brownish, becoming paler near junction to thorax; closed cell M in hindwing present; tarsal claws with conspicuous subbasal tooth.
Head. Area between frontal crests reaching or slightly exceeding the level of crests in dorsal view; malar space less than or equal to distance between antennal sockets; length of postocellar area about 2 times of lateral ocellus diameter; maximal length of temple 1.25-1.45 times greater than its minimal length; flagellum 2.3-2.6 times longer than breadth of head.
Thorax. Distance between cenchri variable, up to 2 times width of cenchrus. Tarsal claws with minute subbasal tooth.
Taxonomic affinities. Based on the similarities in penis valves, the closest species is E. sulcata Wei & Nie, 1998 from China (see http://www.morphbank. net/?id=643394). While the penis valves of both species can easily be distinguished, the distinctly concave dorsal margin of valviceps of these species is a unique characteristic within Empria. Serrulae of the two species are clearly different (cf. Fig.  21  Taxonomic affinities. The most similar species morphologically appears to be Nearctic E. ignota (Norton, 1867). The clearest differences between these species can be seen in the structure of penis valves ( Fig. 32; http://www.morphbank.net/?id=694564).
Taxonomic affinities. The closest species are E. zhangi Wei & Yan, 2009 (China) and E. rubicola Ermolenko, 1971. Empria zhangi (two females and two males studied, including the holotype) can be distinguished from E. quadrimaculata mainly by the following two characters: 1) in female malar space clearly less than two times of the lateral ocellus diameter (about two times in E. quadrimaculata and E. rubicola), in male equal or slightly less than the ocellus diameter (clearly longer in E. quadrimaculata and E. rubicola); and 2) in female flagellum about 2.0 times longer than breadth of head (2.1-2.5 times in E. quadrimaculata), in male 2.4-2.5 times (2.9-3.3 times in E. quadrimaculata). Empria rubicola has shorter antennae and three pairs of pale patches (mostly two in E. quadrimaculata) on terga. The penis valves of E. zhangi and E. quadrimaculata are very similar (http://www. morphbank.net/?id=693502; Fig. 26), while E. rubicola can be distinguished from the two by relatively large basal lobe of the valviceps and by the ventroapical part clearly bent towards its basal part (Fig. 27). Valvula 1 appears indistinguishable in all three species. Host plants. Okutani (1954) indicated Geum japonicum Thunb., but noted later that the specific identity of the reared Empria species was uncertain (Okutani 1967).
Distribution. East Palaearctic. Specimens studied are from Japan (Hokkaido) and Russia (Sakhalin Oblast). Most probably this species has to be removed from the list of Chinese species (Yan et al. 2009), because E. rubicola has clypeus and upper half of the mesepisternum black (not yellow brown) and abdominal terga 2-4 (not 2-6) each with a pair of pale patches. Female. Body length. (5.1)6.4-6.9 mm. Colour. Black; following parts more or less unpigmented, whitish or yellowish brown: labrum; apical maxillary and labial palpomeres; tegulae completely; posterodorsal margin of pronotum in lateral part rather widely, upper part of posterolateral margin of pronotum quite narrowly; pro-, meso-, and metacoxa apically; pro-, meso-, and metatrochanter partly or in most part; pro-, meso-, and metatrochantellus partly or completely; profemur in anterior, posterior, and lateral aspects; mesofemur and metafemur apically slightly; protibia in anterior and posterior aspects; mesotibia in most part; metatibia in basal 2/3; tarsomere 1 of hind leg in basal 2/3; paired patches on abdominal terga 2-4(5); posterior margins of terga and sterna; and cenchri (in one female only posterior margin).

Empria takeuchii
Head. Head behind eyes in dorsal view subparallel sided; postocellar area trapeziform, its length mostly less than or equal to 2 times of lateral ocellus diameter; area between frontal crests in dorsal view reaches or slightly exceeds the level of crests; face and clypeus with somewhat irregular punctures, less shining compared to vertex and especially to postocellar area; ocellar and postocellar area at least slightly raised; clypeus tridentate, with median tooth smaller than lateral teeth; clypeus with median keel; malar space (minimal ventro-ocular distance) shorter or equal to distance between antennal sockets; frontal ridge "V"-shaped, central part of frontal field with distinct pit; maximal length of temple 1.25-1.4 times greater than its minimal length; flagellum 1.8-2.0 times longer than breadth of head.
Thorax. Anterior part of mesoscutum with more or less distinct punctures, its median and postero-lateral portions in most part with sparse indistinct punctures and glossy interspaces, or almost impunctate, glossy; mesoscutellum, mesoscutellar appendage, and metapostnotum impunctate and glossy; mesepisternum with more or less indistinct punctures, mostly glossy; mesepimeron with setae on posterior part; metepisternum with evenly distributed setae; metepimeron in central part without setae; distance between cenchri in most specimens about equal to cenchrus width, but sometimes slightly greater; wings hyaline with brownish venation; closed cell M in hindwing present; tarsal claws with conspicuous subbasal tooth.
Male. (Mostly the differences compared to female are given). Body length. 5.6-5.8 mm.
Head. Area between frontal crests in dorsal view not exceeding the level of crests; length of postocellar area 1.5-2.0 times of lateral ocellus diameter; maximal length of temple 1.25-1.45 times greater than its minimal length; flagellum 2.2-2.7 times longer than breadth of head.
Taxonomic affinities. Morphologically, no certain closest relative can be specified. Superficially may resemble E. rubicola (based on males), E. honshuana (based on females), or E. tridentis (both have pale trochanters and trochantelli). Penis valve (Fig. 30) and valvula 1 (Fig. 19) clearly distinguish this species from all other known species of Empria. According to the molecular analyses (of ITS1 and ITS2 combined with mtDNA sequences), the closest species are those of the E. longicornis and E. immersa species groups, and E. tridentis (Figs 38, 40).
Host plants. Unknown. Distribution. Japan (Hokkaido, Honshu). Etymology. The specific name refers to Kichizo Takeuchi (1892-1968, who made great contributions to the sawfly systematics in eastern Asia. Notes. Six additional studied specimens (1 female, 5 males) from Hokkaido were not included in the type series. The female and most of the males have a longer postocellar area (more than 2 times of the lateral ocellus diameter) compared to the specimens from Honshu (mostly less than 2 times). Serrulae of the Hokkaido female are also slightly different (cf. http://www.morphbank.net/?id=693521 and Fig. 19). No clear differences were found in the structure of penis valves between the specimens from Hokkaido and Honshu.  Dovnar-Zapolskij, 1929: 38-39. Synonymy according to Conde (1940), see Prous et al. (2011b) for details. Empria (Triempria) konowi Dovnar-Zapolskij, 1929: 39-40 Taxonomic affinities. Morphologically, no close relatives can be identified, but in the phylogenetic analysis of the ITS and mtDNA sequences combined, the species appears as a sister of the longicornis-group (Fig. 40) Notes. The original description of this species states that there are "a pair of large flecks on lateral portion of lst-4th tergite" (Lee and Ryu 1996), while actually no specimen studied (including the holotype) has pale patches ("large flecks") on first tergite. There is one male (NSMT018) from Honshu (Nagano) with penis valve slightly different (see http://www.morphbank.net/?id=592669) from all the other studied males, but the material is currently insufficient to decide if the specimen is aberrant or represents a separate (sibling) species.

Empria sp. 1
Taxonomic affinities. Belongs to E. longicornis group. Externally it is most similar to E. japonica, but penis valve is clearly distinct from all other known species of the longicornis-group (Fig. 36), being most similar to E. alpina Benson, 1938 (e.g. http://www. morphbank.net/?id=577439). Can be distinguished from E. alpina by its colouration: in E. sp1 tegulae, posterior margin of pronotum, and basal 1/3 of metatibia are pale, while in E. alpina these are mostly black. Distinctness of this taxon is also supported by nuclear ITS sequence data (Fig. 38).
Host plants. Unknown. Distribution. Japan (Hokkaido). Notes. Because taxonomy of the longicornis-group is quite difficult (Prous et al. 2011b) and the corresponding female remains to be found yet, additional material is needed to describe and name this presumably new species.

Molecular phylogenetic analyses
Bayesian analyses of the mitochondrial and nuclear sequences separately and in combination all resulted in somewhat different topologies , with well supported differences in some cases (especially in the longicornis and the immersagroups). However, several clades were reconstructed in all analyses with significant statistical support (posterior probability 0.95 or more). Based on these analyses, the basal split within the genus Empria is between E. candidata and all other species (Figs 38-40), which is consistent with the division of the genus into two subgenera, Parataxonus MacGillivray, 1908 (E. candidata) and Empria s. str. (Zhelochovtsev and Zinovjev 1988;Yan et al. 2009). Monophyly of the immersa-group, the longicornis-group, and the quadrimaculata-group is well supported in all our analyses (Figs 38-40). Empria quadrimaculata species group is proposed here for the first time for the species sharing the same type of lancets (Figs 15-16; http://www. morphbank.net/?id=693500) and penis valves (Figs 26-27; http://www.morphbank. net/?id=693502). A clade comprising the longicornis-group and the immersa-group, E. tridentis, and E. takeuchii is well supported in the analysis of nuclear ITS and in the combined analysis of ITS and the mitochondrial sequences (Figs 38,40). In the analysis of the mitochondrial DNA sequences, however, E. takeuchii is excluded from this clade, but without significant support for any other sister-group relationships within Empria s. str. (Fig. 39). The sister group of E. honshuana, revealed in the analyses of ITS and the combined sequences, is E. pallimacula (Figs 38, 40), but according to the mitochondrial sequences, it is E. excisa (Fig. 39).
Each of E. japonica, E. loktini, E. longicornis, E. immersa, and E. plana is monophyletic (as would be expected from morphology) according to the ITS sequences (Fig.  38), but not according to the mitochondrial DNA (Fig. 39). The monophyly of Empria tridens is supported neither by ITS nor the mitochondrial sequences (Figs 38-39; see discussion in Prous et al. 2011b). Remarkably, Empria sp. 1 (USNM2051678_040) has an identical mitochondrial haplotype with one specimen of E. loktini (TUZ615180), while morphology (cf. Figs 33 and 36, see also the key) and the nuclear ITS sequences (Fig. 38) clearly differentiate these species. Because the four independent runs of BAli-Phy produced different topologies, only clades which were found in all trees and were supported with posterior probabilities (PP) 0.9 or more are shown. Duplicate (shown behind the sequence used in the analysis) and very similar sequences (three E. japonica, two E. tridentis, and one E. rubicola) were removed prior to analyses to reduce computation time. 39 Phylogeny of mitochondrial sequences using MrBayes (GTR + I + G[4] model; alignment length 1642 bp). Duplicate sequences (shown behind the sequence used in the analysis) were removed prior to analyses. Empria liturata from Japan (USNM2051678_021) was also excluded due to incomplete sequence. 40 Combined analysis of ITS (MAP alignment from BAli-Phy analysis) and mitochondrial sequences using MrBayes (GTR + I + G[4] model). Monsoma pulveratum was used as an outgroup. Clades with posterior probabilities (PP) less than 0.9 were collapsed in all the trees.

Discussion
Although identification of Empria species using only external morphology can often be difficult, we found that females of the species reviewed here can mostly be identified without dissecting their ovipositors. Identification of the males is much less reliable without studying their genitalia because of more extensive intraspecific variation and less pronounced differences among species. The most difficult species to separate from each other on the basis of female characters are E. quadrimaculata and E. rubicola, the ovipositors of which appear nearly indistinguishable (Figs 15-16). Also the external characters applied in the present key overlap considerably between them. However, because there are consistent differences in the penis valves between the two (see Figs 26-27), they most likely represent different species.
Due to the general difficulty in identifying the Empria species using only external morphology, it is advisable in our opinion to leave the specimens unidentified (to avoid possible confusions in the future), especially those from the poorly studied regions (e.g. Eastern and Central Asia), as long as their identity remains problematic from external morphology and the genitalia cannot be dissected.
In addition to the 11 named Empria species and one presumably new but undescribed species (currently only one male is known) reported here, some additional species of the genus are likely to be found in Japan. Alpine habitats above the tree line might be inhabited by additional Empria species, but from there we have no samples yet.
The results of our molecular phylogenetic analyses (Figs 38-40) significantly supported the groupings within Empria that could be expected from morphology (Empria s. str., immersa-group, longicornis-group, and quadrimaculata-group). Although E. pumiloides was the only species from the hungarica-group in the current dataset, monophyly of this group is also supported by DNA data (unpublished results). The consistent affinity found between the longicornis-group, the immersa-group, and E. tridentis in all our analyses  was the only phylogenetic result not expected from morphology (though phylogenetic analyses using morphological data are still lacking). Based on the phylogenetic results presented here, we cannot draw any more definite conclusions regarding the phylogeny of Empria, which require, in addition to improving taxon and gene sampling, possibly also methodological advancements (e.g. using methods which take into account incomplete lineage sorting; Heled and Drummond 2010). The conflict between ITS and mitochondrial phylogenies within the E. longicornis and the E. immersa species groups (Figs 38-39; see also Prous et al. 2011b) needs further study as well (e.g. sequencing 1-3 additional nuclear markers). However, we note that incongruence between mitochondrial phylogeny with morphology and nuclear phylogeny is not uncommon among closely related species, possibly because of mitochondrial introgression (e.g. Linnen and Farrell 2007;Wahlberg et al. 2009;Near et al. 2011). Another explanation, which we cannot exclude based on current data, might be incomplete lineage sorting (for a review, see Degnan and Rosenberg 2009).