To photograph penis valves and lancets (valvula 1 or ventral part of saw), genital capsules and ovipositors were separated from the specimen and macerated in KOH (10–15%) for 6–10 hours at room temperature or treated with proteinase during DNA extraction (see below). Temporary or permanent slide preparations were made of dissected lancets and penis valves. For temporary slides, glycerine was used. After photographing, the lancets and penis valves were glued on a piece of cardboard, which was pinned with the corresponding specimen. For permanent slides, Euparal or PVA-mounting medium (Danielsson 1985) was used (these specimens are labelled as ‘PR.XXXVV’, e.g. PR.579VV). Photos were taken with a digital camera attached to a microscope. Composite images with an extended depth of field were created from stacks of images using the software CombineZP (Alan Hadley; http://www.hadleyweb.pwp.blueyonder.co.uk/). Most of the lancets were photographed in two overlapping parts and a single image created using the program Image Composite Editor (Microsoft), or in few cases with plugin MosaicJ (Thévenaz and Unser 2007) implemented in ImageJ version 1.46r (Wayne Rasband; http://imagej.nih.gov/ij/). Morphological terminology follows Vikberg (1978, 2006) and Viitasaari (2002). We do not use the term “ctenidia” to denote the bands of setae on the annuli of female lancets (Fig. 129), as used previously by Prous et al. (2016) following Viitasaari (2002, Fig. 134). The term “ctenidia” (singular “ctenidium”) correctly refers to rows of spines (i.e. protrusions of the cuticle); superficially they resemble bands of setae, which are composed of sensilla (Smith 1968, Blank and Schönitzer 1994, fig. 131 in Viitasaari 2002). Therefore we use the more neutral term “setae”.

All the rearing data newly published here were obtained by VV. These data are given under Rearing notes in the Taxonomy section. Plants chosen for ovipositing experiments were mainly those from which females were collected in the field or reared from larvae. Our main intention was to find at least one acceptable host plant for rearing particular species. Ovipositing experiments were carried out indoors, at room temperature. Suitable parts of plants were cut, the cut end put in water or wrapped in moistened filter paper (paper towel), and the leaves offered to the sawfly in a closed container. Larvae were fed with fresh leaves that were changed every day or every second day. Cocoons were placed in glass vials and overwintered outdoors in a wooden chest.

DNA was extracted and purified with an EZNA Tissue DNA Kit (Omega Bio-tek) according to the manufacturer’s protocol and stored at -20 °C for later use. Typically, the middle right leg was used for DNA extraction, but for males the whole genital capsule was often additionally used to increase DNA yield and to free penis valves from muscles for photographing. One mitochondrial and two nuclear regions were used in phylogenetic analyses. Primers used for amplification and sequencing are listed in Table 1. The mitochondrial region used is a large fragment (1078 bp) of cytochrome oxidase subunit I gene (COI). The first (from the 5’ end) 658 bp of this fragment correspond to the standard barcode region of the animal kingdom (Hebert et al. 2003). If the amplification of the 1078 bp fragment failed, or was expected to fail because of low DNA quality, the region was amplified in two overlapping fragments, or only the barcoding (658 bp) region was obtained. The nuclear markers used are fragments of triose-phosphate isomerase (TPI) and sodium/potassium-transporting ATPase subunit alpha (NaK). The TPI fragment used is the nearly complete gene region, containing 676 bp of three exons and two short introns (each around 50–100 bp) in Nematinae, altogether 795–833 bp. The NaK fragment used is a nearly complete sequence of its longest exon, 1654 bp. New NaK primers were designed mainly based on four sawfly genomes and one transcriptome available in GenBank (accessions AOFN01001568, GAWW02019159, LGIB01000323, AMWH01001469, AZGP01005167), or using sequences published by Malm and Nyman (2015). Numbers in the new NaK primer names refer to the binding position of the primer’s 3’ end in the coding region of Athalia rosae mRNA (accession XM_012414227). Seven sequences of four specimens (4b, 9t, DG, DH) newly reported here were obtained from the VoSeq database (Peña and Malm 2012) maintained by Tommi Nyman (University of Eastern Finland). For these four specimens, 810 bp of COI and 997 bp of NaK had been sequenced as described previously (Nyman et al. 2006, Leppänen et al. 2012).

PCR reactions were carried out in a total volume of 15–25 μl containing 1–2 μl of extracted DNA, 1.0–1.5 μl (5.0–7.5 pmol) of primers and 7.5–12.5 μl of 2x Multiplex PCR Plus Master mix (QIAGEN). The PCR protocol consisted of an initial DNA polymerase (HotStar Taq) activation step at 95 °C for 5 min, followed by 38–40 cycles of 30 s at 95 °C, 90 s at 47–59 °C depending on the primer set used, and 30–120 s (depending on the amplicon size) at 72 °C; the last cycle was followed by a final 30 min extension step at 68 °C. 3 μl of PCR product was visualised on a 1.4% agarose gel and then purified with FastAP and Exonuclease I (Thermo Scientific). 1.0–1.8 U of both enzymes were added to 12–22 μl of PCR solution and incubated for 15 min at 37 °C, followed by 15 min at 85 °C. Purified PCR products were sent to Macrogen (Netherlands) for sequencing. To obtain unequivocal sequences, both sense and antisense strands were sequenced, using the primers listed in Table 1. Some TPI sequences were polymorphic for intron length and in those cases heterozygous insertions/deletions (indels) were reconstructed using the program Indelligent v.1.2 (Dmitriev and Rakitov 2008), available at http://dmitriev.speciesfile.org/indel.asp. Only the longest haplotype was submitted to Genbank, because Genbank does not accept more than one sequence per specimen and marker. Ambiguous positions (i.e. double peaks in chromatograms of both strands) due to heterozygosity or heteroplasmy were coded using IUPAC symbols. The COI sequence of one sequenced specimen (Pristiphora friesei DEI-GISHym11558) had a single bp deletion, which was replaced with N before submitting to GenBank.

Sequences reported here have been deposited in the GenBank (NCBI) database (accession numbers KY698031–KY698412 and MF426916–MF426924). Some of the sequences analysed here were originally published by Nyman et al. (2006, 2010), Prous et al. (2014, 2016), and Schmidt et al. (2017).

Sequences of COI, NaK, and TPI without introns were aligned manually, among which only one COI sequence of P. friesei (DEI-GISHym11558) had a single bp deletion.

For concatenation of separate alignments of different genes we used FASconCAT-G (Kück and Meusemann 2010) (https://www.zfmk.de/en/research/research-centres-and-groups/fasconcat-g). To minimize the missing cells in the dataset, a few composite COI sequences were created from different specimens of the same species. A nearly complete 1078 bp region of COI was created in these cases by combining the barcoding region (up to 658 bp) with the 810 bp region (423 bp of which overlaps with the barcoding region) when the overlapping part was identical between the sequences. Both ID numbers of the specimens used to create the composite COI sequences are given in the figures.

For phylogenetic analyses we used the maximum likelihood method (ML) implemented in RAxML v. 8.2.9 (Stamatakis 2014) through CIPRES Science Gateway V. 3.3 (Miller et al. 2010) at https://www.phylo.org/. Robustness of reconstructed trees was estimated with 1000 rapid bootstrap replicates. For tree search, GTRCAT model with 25 site-specific rate categories was used and the final tree evaluated with the GTRGAMMA model (Stamatakis 2006; http://sco.h-its.org/exelixis/web/software/raxml/). The concatenated dataset was partitioned by genes, and model parameters estimated separately for each gene. Additionally, MEGA7 (Kumar et al. 2016) was used to calculate p-distances (proportion of nucleotide differences) between specimens and net synonymous divergence between species. Unless stated otherwise, TPI introns were excluded from calculations. Minimal p-distances between and maximal distances within BIN (Barcode Index Number) clusters were taken from BOLD (http://www.boldsystems.org/) BIN database. Alignment files and output files from phylogenetic analysesare available at figshare (http://dx.doi.org/10.6084/m9.figshare.5235832).

Some of the COI barcode sequences used here were obtained from BOLD (http://www.boldsystems.org/). In this case, DNA extraction, PCR amplification, and sequencing were conducted at the Canadian Centre for DNA Barcoding (CCDB) in Guelph, Canada, using standardised high-throughput protocols (Ivanova et al. 2006, deWaard et al. 2008), available online under www.ccdb.ca/resources.php. DNA aliquots of SDEI vouchers are deposited in the DNA storage facility of the SDEI (including those that were originally extracted in CCDB).

The electronic identification key for the species of Pristiphora was prepared in Lucid 3.5 Builder (http://www.lucidcentral.org/) and a zip file containing all the Lucid data files is available at figshare (http://dx.doi.org/10.6084/m9.figshare.5235805). If the licence for Lucid 3.5 is lacking, the free version of Lucid 3.3 can be used to run the key. In case of ambiguities or polymorphisms in character states, we conservatively coded these as multiple states. The key contains 90 morphological features with 244 character states, and 68 male and 71 female entities (species or groups). The first choice given in the key is between female and male, one of which has to be chosen to see all other characters. After that, characters can be chosen freely or one can use ‘Best’ and ‘Next Best’ tools in Lucid to suggest the most efficient sequence of characters for identification. A traditional dichotomous key was constructed manually to emphasise the most reliable characters (usually penis valve for males, and valvula 3 or lancet for females).

The genus was recently delimited mainly by phylogenetic analyses of DNA sequence data by Prous et al. (2014), which is here confirmed with the addition of the nuclear gene TPI. Phylogenetic analysis of mitochondrial COI and nuclear NaK and TPI sequences (altogether 3408 bp) strongly support monophyly of Pristiphora (Fig. 1). There are no unambiguous morphological characters that define Pristiphora, but the combination of the following characters can be used to distinguish most of the species from the similar genera Nematus and Euura (see Prous et al. 2014): clypeus usually more or less truncate (Fig. 9); apex of vein C usually swollen; head length behind eyes usually small; claws often with small subapical tooth or simple; in females, valvula 3 (apical sawsheath) often with scopa (e.g. Figs 75, 86, 104–107, 127) and tangium of lancet nearly always with campaniform sensilla (Fig. 177); in males, posterior end of tergum 8 without distinct apical projection in most species (except P. armata and P. leucopus), and penis valve usually with Valvispina arising close to the ventral margin. Vikberg (1982) lists larval characters that might more reliably distinguish Pristiphora from Nematus and Euura with morphologically similar adults: 3rd abdominal segment with 6 annulets, annulets 2 and 4 with setae; no cerci; and maxillary stipes with or without seta. However, not all of these characters apply to all Pristiphora. For example, larvae of P. dasiphorae have only 3 visible annulets on abdominal segments and both species of the P. malaisei group (arctica group in Prous et al. 2014), P. malaisei and P. dasiphorae, have setae also on annulet 1 (Zinovjev 1993). More Pristiphora species should be scored for the larval characters mentioned by Vikberg (1982) to test the reliability of these characters for defining the genus.

Some of the nuclear sequences from males included polymorphic sites (double peaks in chromatograms) indicating heterozygosity or the presence of paralogous genes. Because these few polymorphic sites are restricted to synonymous positions (i.e. not affecting protein sequence) and so far there is no evidence for paralogous NaK and TPI genes at least in sawflies, these polymorphic sites might indicate heterozygosity because of diploidy. This can happen if there is no heterozygosity at the complementary sex determination (CSD) locus, preventing female development from a diploid embryo (Naito and Suzuki 1991, Heimpel and de Boer 2008, Harper et al. 2016). Survival of diploid males is known in sawflies (Naito and Suzuki 1991, Cook et al. 2013, Harper et al. 2016) and is common in nature at least in some species of Apocrita (Liebert et al. 2005, Retamal et al. 2016).

Figure 1. 

Maximum likelihood tree of Pristiphora and nematine outgroups based on three genes (3408 bp). Specimens having at least two of the three genes were included. Numbers above branches show bootstrap proportions (%). Support values for weakly supported branches (BP<70) are not shown. Part of Pristiphora is shown without support values and tip labels, which are shown in Fig. 2. The scale bar shows the number of estimated substitutions per nucleotide position.

COI sequence of one P. friesei specimen (DEI-GISHym11558) had a single bp deletion, which was caused by shortening of 8 bp repeat of thymines (T) to 7 thymines, compared to a different P. friesei specimen and closely related P. laricis. Because the P. friesei sequence with the deletion was of good quality (trace files at figshare http://dx.doi.org/10.6084/m9.figshare.5235832), was found in amplicons obtained with two sets of different primers, and did not show anomalous phylogenetic position, it is possible that this COI sequence is functional despite the frameshift mutation (Beckenbach et al. 2005, Rosengarten et al. 2008). Barcode sequences of two additional P. friesei specimens in BOLD that are labelled as NUMT (nuclear mitochondrial pseudogene), DEI-GISHym4993 and DEI-GISHym11557, have identical sequence to specimen DEI-GISHym11558.

Phylogenetic relationships at the base of Pristiphora are not well resolved by our three-gene dataset (Fig. 1), because most of the deeper splits receive bootstrap support less than 70%. Many of the more derived clades that receive strong bootstrap support (more than 90%) are also morphologically well-supported (see group definitions below). Generally, our results strengthen the conclusions of Nyman et al. (2010) that only a minority of lineage splits correlate with host plant shifts. Species in the groups which monophyly is well supported (Figs 1–2) tend to share the same host plant genus or family (see group definitions below and Nyman et al. 2010 for details). We will not repeat the quantitive analysis of Nyman et al. (2010), but our broader taxon sampling yields additional support for their conclusions, e.g. we recover a Quercus-feeding clade composed of P. fausta, P. calliprina, and P. parnasia (Figs 2, 5), and an Acer-feeding depressa group (Figs 1, 3), although host plants are not known for all the species involved.

To enable easier discussion of relationships among Pristiphora species, we define the following informal species groups based on morphology and the phylogenetic analysis: abietina, alpestris, carinata, depressa, erichsonii, laricis, leucopodia, malaisei, micronematica, nigella, pallida, pallidiventris, retusa, ruficornis, and rufipes group. The names of the groups are based on the oldest valid species name in that group.

The following treated species were not assigned to any species groups: abbreviata, angulata, biscalis, bufo, cadma, cincta, condei, conjugata, dedeara sp. n., fausta, geniculata, insularis, maesta, mollis, monogyniae, paralella, pseudogeniculata, punctifrons, tenuiserra, and testacea.

Figure 2. 

Part of the maximum likelihood tree shown in Fig. 1. Numbers above branches show bootstrap proportions (%). Support values for weakly supported branches (BP<70) are not shown. Outline of the full tree is shown upper left, with the part shown here highlighted. The scale bar shows the number of estimated substitutions per nucleotide position.

Figure 3. 

Maximum likelihood tree of Pristiphora based on COI gene (1078 bp). Shortest sequence included was 462 bp (subarctica MHV00166). Numbers above branches show bootstrap proportions (%). Support values for weakly supported branches (BP<70) are not shown. BIN numbers (BOLD:XXXXXXX) referred to in the text are shown for representative specimens. Part of Pristiphora is shown without support values and tip labels, which are shown in Figs 4–5. The scale bar shows the number of estimated substitutions per nucleotide position.

Includes the following North-Western Palaearctic species: abietina, compressa, decipiens, gerula, pseudodecipiens, robusta, and saxesenii. All species feed on Picea (Beneš and Krístek 1979) and the females have a characteristically modified valvula 3 (Figs 116–118, 120–122), which is apparently correlated with oviposition in narrow conifer needles. Specifically, valvula 3 is relatively narrow in dorsal view, and in lateral view its dorsal and ventral margins are more or less parallel, with the posterior margin distinctly truncate (Figs 116–118, 120–122). In addition, the apical part of the abdomen of females is laterally compressed (Fig. 60), but this character is not reliable for unambiguous distinction of females of the abietina group from other species. Males are externally very similar to many other species outside the abietina group that are ventrally extensively pale and dorsally black, so that penis valves should be studied to recognise them. Penis valves of the abietina group have a similar structure (Figs 283–289) that can be distinguished from other Pristiphora: more or less straight ventral margin, similarly sized paravalva and pseudoceps with roughly convex dorsal margin, apically unmodified pseudoceps, and straight or slightly bent valvispina. Genetic data (no data yet for P. robusta) strongly support the monophyly of the abietina group (Figs 2, 5). We exclude the erichsonii and pallida groups from the abietina group (included in Wong 1975), because current phylogenetic analyses do not support such a clade (Figs 2, 5). Unusually, the divergence in mitochondrial sequences (maximum distance 2.2%, minimum between species distance 0.0%) in the abietina group is similar or lower than in nuclear sequences (maximum distance 2.7%, minimum between species distance 1.1%). This is a reversal of the typical condition in bilateral animals (Bilateria), in which the substitution rate in mitochondrial DNA is several-fold higher than in nuclear DNA (Ballard and Whitlock 2004, Lavrov 2007). It is unlikely that mitochondrial DNA (mtDNA) evolves more slowly than nuclear DNA in the abietina group (no reliable cases are known in Bilateria). More probable is that the reduced diversity of mtDNA is caused, for example, by infection with maternally transmitted Wolbachia (or some other Bacteria), a common occurrence in arthropods (e.g. Hurst and Jiggins 2005, Werren et al. 2008). Wolbachia infection can cause cytoplasmic incompatibility between infected males and uninfected females, meaning that mitochondria carried by infected females are preferentially spread within population or between closely related species via introgression. This could explain sharing of (near) identical COI sequences between different species in abietina group.

Includes the following North-Western Palaearctic species: alpestris, pseudocoactula, and possibly dissimilis. Penis valves and lancets are very similar in alpestris and pseudocoactula (Figs 203–204, 233–236), but females are not known for dissimilis and its penis valve is very distinctive compared to all other Pristiphora species (Fig. 237). Genetic data strongly supports monophyly of the alpestris and pseudocoactula clade (Figs 2, 5). Placement of dissimilis (no genetic data yet) in the alpestris group is tentative (Lindqvist 1971).

Includes the following North-Western Palaearctic species: albilabris, borea, breadalbanensis, carinata, coactula, groenblomi, lativentris, and trochanterica. Species of the carinata group have a distinctly matt mesepisternum, and completely or largely black body. Valvula 3 of females varies from simple (short, slightly tapering from base to apex, and without scopa; Figs 98–99) to somewhat square-shaped in dorsal view and with small scopa (Figs 93–97). Lancets are without setae and very similar to each other (Figs 210–217). Penis valves are also hardly distinguishable (Figs 238–242). Because of the high degree of similarity and the apparently continuous variation of (nearly) all characters, it is not clear how many species should be recognised and how they should be delimited. Genetic data strongly supports monophyly of the carinata group (Figs 2, 5). More research is need to associate males and females of different species, preferably by rearing experiments and sequencing of different nuclear genes, so that morphological variation within and between the species can be assessed more reliably.

Includes the following North-Western Palaearctic species: depressa, subbifida, and tetrica. This species group was reviewed by Liston and Späth (2008) and Liston et al. (2013). As far as is known, all species feed on Acer (Liston and Späth 2008). Females have usually a relatively short and narrow valvula 3 with a small scopa (Figs 82–83), but sometimes the scopa is practically absent. Most species have bifid or subbifid (Figs 24–25) claws. Males have a long and thin valvispina that can be strongly bent (Fig. 301), or nearly straight (in P. ifranensis Lacourt, 1973). Combined analysis of mitocondrial and nuclear genes moderately supports monophyly of the depressa group (Fig. 1). Genetic data (Figs 1, 3) indicates several additional, possibly undescribed, species in Europe (Mediterranean region and Central Europe; DEI-GISHym20783, 20784) and in the Far East (DEI-GISHym86127, 80227, 80233).

Includes the following North-Western Palaearctic species: erichsonii, glauca, and wesmaeli. All species feed on Larix (Wong 1975). Characteristically for conifer-feeders, species of the erichsonii group have a narrow valvula 3 (Figs 124, 126). P. erichsonii has a different coloration (abdomen with a red band) and valvula 3 (not apically abruptly constricted; Fig. 124) from P. glauca and P. wesmaeli (abdomen ventrally yellow, valvula 3 apically abruptly constricted Fig. 126), but penis valves (Figs 290–292) and genetic data (Figs 2, 5) support monophyly of this group. The penis valves of the erichsonii group have a somewhat rhombus- (Fig. 290) or trapez-shaped (Figs 291–292) paravalva, a rather rectangular and unmodified pseudoceps, and small and slightly bent valvispina that is asymmetrical at apex.

Includes the following North-Western Palaearctic species: friesei and laricis. Both species feed on Larix (Adam 1973, Huflejt and Sawoniewicz 1999, Liston et al. 2006). Females have a short valvula 3 (cerci clearly extend beyond its apex), with small scopa (Figs 79–81). The absence of a velum on the anterior protibial spur (Fig. 33), and usually mostly black body, can help in recognising this group. Lancets (Figs 174–178) and penis valves (Figs 277, 279–208) unambiguously distinguish the laricis group from other species (see the Key). Genetic data strongly supports monophyly of the laricis group (Figs 1, 3).

Includes the following North-Western Palaearctic species: leucopodia and nigriceps. Both species feed on Picea (Beneš and Krístek 1979). As in many other species feeding on conifers, valvula 3 is narrow in the leucopodia group (Fig. 84). In lateral view, valvula 3 has a round outgrowth at the posterior margin (Fig. 85), which distinguishes the leucopodia group from other conifer-feeders. A relatively similar valvula 3 is found in P. insularis (Figs 86–87), but other characters distinguish this species from the leucopodia group (anterior protibial spur with velum and clearly different lancet; see Figs 33–34, 135, 181–183). A characteristic hump that is posteriorly constricted and situated at the base of the slightly bent valvispina (Figs 281–282) distinguish penis valves of the leucopodia group from other Pristiphora (see the Key). Combined mitochondrial and nuclear phylogenetic analysis strongly supports monophyly of the leucopodia group (Fig. 1), but mitochondrial COI sequences alone do not (Fig. 3).

Includes the following North-Western Palaearctic species: malaisei and dasiphorae. Both species feed on Potentilla fruticosa L. (=Dasiphora fruticosa) (Zinovjev 1993) and possibly Comarum palustre L. Quite characteristic for the group is its head shape, which is more evident in females (Fig. 58). Structure of the penis valves (apically truncate and without valvispina; Figs 297–298) is very distinct within Pristiphora, prompting Zinovjev (1993) to create even a separate tribe for the species of malaisei group. Genetic data strongly supports monophyly of the group (Fig. 1).

Includes the following North-Western Palaearctic species: affinis, atripes, kontuniemii, micronematica, nordmani, reuteri, sermola, and possibly lanifica. In contrast to other group names, we use the name micronematica (Malaise 1931) for this group instead of the oldest name lanifica (Zaddach in Brischke 1883), because the identity of lanifica is uncertain and it might not belong to this group. As far as is known, species of this group feed on Salix (Lindqvist 1952, Kontuniemi 1960, 1972, Vikberg 1966, Liston 1982, Kangas 1985). Valvula 3 in females has a small scopa (Fig. 92), is more or less square-shaped in dorsal view and has two dense, lateral bundles of hairs at its apex (Fig. 91). Very similar are species in the carinata and alpestris groups, but lancets (with setae and serrulae not papilliform; Figs 205–209) and penis valves (pseudoceps in most species dorsally with loose membranous region covered with hair; Figs 303–310) of the micronematica group can be readily separated from these groups (see the Key). Genetic data strongly supports monophyly of the micronematica group (Figs 2, 5). Ecologically, and in some other biological characteristics, this group is similar to the phylogenetically distant aphantoneura subgroup (Figs 2, 4; Prous et al. 2016): the majority of species (perhaps all in the micronematica group) feed on Salix, are more abundant and species rich in boreal or (sub)arctic habitats (probably because of the host plants), have low diversity in mitochondrial DNA (maximum divergence in COI barcodes is 2.5% in micronematica group, 3.3% in aphantoneura subgroup) that does not correlate with species boundaries (tentative for micronematica group), and have phenotypically exceedingly similar females, while males can in most cases be identified by clear differences in penis valves. At least ex ovo rearings or sequencing of several nuclear markers from most species is needed to associate males and females confidently in the micronematica group.

Includes the following North-Western Palaearctic species: amphibola, nigella, and parva. All species feed on Picea (Beneš et al. 1981). The structure of valvula 3 in the nigella group is unique among Pristiphora, with the scopa positioned dorsoapically (Figs 111, 115), rather than just apically or ventroapically as in other species. This might be related to adaptation to ovipositing into closed or opening Picea buds (Nägeli 1936, Benson 1948, Grebenshchikova 1986), in contrast to other conifer-feeders which use already expanded needles. Lancets (long and narrow, with setae and flat serrulae; Figs 165–167) and penis valves (apically narrowed pseudoceps, paravalva dorsally somewhat s-shaped and with small dorsally directed valvispina; Figs 293–295) of the nigella group can also be distinguished from other species. Genetic data strongly supports monophyly of the nigella group (Figs 2, 5).

Figure 4. 

Part of the maximum likelihood tree shown in Fig. 3. Numbers above branches show bootstrap proportions (%). Support values for weakly supported branches (BP<70) are not shown. BIN numbers (BOLD:XXXXXXX) referred to in the text are shown for representative specimens. Outline of the full tree is shown upper left, with the part shown here highlighted. The scale bar shows the number of estimated substitutions per nucleotide position.

Figure 5. 

Part of the maximum likelihood tree shown in Fig. 3. Numbers above branches show bootstrap proportions (%). Support values for weakly supported branches (BP<70) are not shown. BIN numbers (BOLD:XXXXXXX) referred to in the text are shown for representative specimens. Outline of the full tree is shown upper left, with the part shown here highlighted. The scale bar shows the number of estimated substitutions per nucleotide position.

Includes the following North-Western Palaearctic species: pallida and subarctica. Both species feed on Picea (Beneš and Krístek 1979). As in many other conifer-feeders, species of the pallida group have a narrow valvula 3 (as in Fig. 124) and both males and females are at least ventrally extensively pale yellow. The structure of valvula 3 cannot be distinguished from P. erichsonii, but this species has a different coloration (abdomen with a red band; as in Fig. 47). Very characteristic penis valves (Figs 276, 278) clearly distinguish this group from other similar conifer-feeders (see the Key). We cannot exclude the possible conspecifity of pallida and subarctica, because the differences between them are morphologically (Figs 171–172, 276, 278) and genetically (Fig. 5) small. Examination of more specimens and additional nuclear gene sequences are needed to decide this.

Includes the following North-Western Palaearctic species: nigricans and pallidiventris. Both species feed on herbaceous Rosaceae (Liston 2011, this study). Usually, in both species the abdomen is at least ventrally pale, but nearly completely black specimens of pallidiventris can occasionally be found. The structure of valvula 3 (deep scopa with long medial projection; Fig. 106) distinguishes this group from other species of Western Palaearctic Pristiphora, though the difference is small compared to many other species with similar valvula 3. When in doubt, lancets should be studied (Figs 141–144). Penis valves of both species are very similar (Fig. 223–226) and can be distinguished from other species: paravalva and pseudoceps are of similar size and shape, ventral margin of paravalva is more or less straight, and the valvispina is straight, small, and positioned in the middle or upper third of paravalva. Current genetic data (Fig. 5) does not allow separation of P. nigricans (only one COI sequence available) and P. pallidiventris.

Includes the following North-Western Palaearctic species: exigua and retusa. A host plant is known only for P. retusa (Prunus padus; Kangas 1985). Characteristic for the group is small size (body length 3.0–4.5 mm), mostly black body, and distinctly asymmetrical labrum (Fig. 11). These two species appear to be rather closely related (only one COI sequence available for P. exigua) (Fig. 3).

Includes the following North-Western Palaearctic species divided into five subgroups (Prous et al. 2016): albitibia subgroup (albitibia, astragali, sootryeni, caraganae), aphantoneura subgroup (aphantoneura, bifida, confusa, luteipes, opaca, pusilla, staudingeri, subopaca), appendiculata subgroup (appendiculata), armata subgroup (armata, leucopus), and ruficornis subgroup (melanocarpa, ruficornis, and possibly frigida). Species of the albitibia subgroup feed on Fabaceae, the aphantoneura subgroup mostly on Salix (except P. aphantoneura that feeds on Lathyrus), P. appendiculata on Ribes, the armata subgroup on Crataegus and Tilia, and the ruficornis subgroup on Betula (a host is not known for P. frigida). Morphologically, the ruficornis group is best characterised by the structure of penis valves, which have a large and usually strongly bent valvispina that almost completely (Figs 263–274) or largely (Fig. 262) replaces the paravalva. Several other species have a similarly large and bent valvispina (Figs 276, 278, 299, 301), but these penis valves can nevertheless be clearly distinguished from those of the ruficornis group, in combination with differences in body coloration (see the Key). Genetic data strongly supports monophyly of the ruficornis group (Figs 2, 4).

Includes the following North-Western Palaearctic species: brevis, dochmocera, rufipes, thalictri, and thalictrivora. The species feed on Aquilegia (P. rufipes) or Thalictrum (P. brevis, P. thalictri, P. thalictrivora, and possibly P. dochmocera). Species of this group are usually black-bodied, valvula 3 of females has a distinct scopa, and lancets are without setae. Lancets and penis valves are very similar in all species. Penis valves of the rufipes group have a somewhat rectangular and unmodified pseudoceps, a somewhat oval-, rectangular- or square-shaped paravalva that is dorsoapically abruptly narrowed before the valvispina, a small and straight valvispina that arises on the ventral part of the paravalva, and a valvar strut that is distinct along its entire length (Figs 243, 245, 247–251). Very similar penis valves (Figs 244, 246), but somewhat more elongate than in the rufipes group, are found in the closely related P. cincta (Figs 2, 5). Species of the rufipes group are also genetically closely related to each other (Figs 2, 5). Species boundaries in the rufipes group are unclear, as well as the association of males and females. At least two species seem to be involved, one feeding on Aquilegia (P. rufipes) and the other one(s) on Thalictrum. The lancet of the Aquilegia-feeding P. rufipes seems to be slightly different from the other species. The most protruding part of the serrulae of rufipes tends to be relatively acute, usually with denticles along the entire margin (Fig. 222), while the most prominent part of the serrulae of the other species tends to be relatively blunt, with denticles concentrated mostly on the apical part of the serrula. Based on mitochondrial and nuclear sequences of larvae collected from Aquilegia sp. (DEI-GISHym20983 and DEI-GISHym21482), we were able to associate three additional adults (DEI-GISHym15263, DEI-GISHym19795, and 9t from VoSeq database) with these larvae. Nuclear sequences from these adults (both genes, only NaK, or only TPI) and the larvae (both genes) form a monophyletic group to the exclusion of other species (not shown). Nuclear data suggests that there might be at least three additional lineages, which we have identified as P. brevis, P. thalictri, and P. thalictrivora (Fig. 2), but morphological separation of these forms is somewhat arbitrary. Although there is some variation in the shape of serrulae of lancets among the Thalictrum-feeders (Figs 218–221), this does not seem to correlate with external morphology, and the variation is continuous. Specimens with completely black (P. thalictri) or more or less completely pale legs (e.g. P. thalictrivora) can both have lancets of the two extreme forms (Figs 218, 221), but apparently also of intermediate form (Figs 219–220). Nevertheless, in addition to P. rufipes, we tentatively recognise four other species: brevis, dochmocera, thalictri, and thalictrivora. Pristiphora thalictri is the darkest, with completely or almost completely black body, including the legs and tegulae. Pristiphora brevis and P. thalictrivora have completely or nearly completely pale legs (at most with black hind tarsi) and tegulae. The only differences between P. brevis and P. thalictrivora may be the longer postocellar area (Fig. 7) and (usually?) rather smooth mesopostnotum (Fig. 19) in P. brevis (matt in P. thalictrivora; Fig. 20). Pristiphora dochmocera, which we tentatively keep as a separate species, might be a synonym of P. thalictri or P. thalictrivora, because its coloration is intermediate between them. Pristiphora dochmocera has black tegulae and pale (metafemur and metatibia nearly completely pale) or relatively dark legs (black metafemur and pale metatibia). Although specimens of the rufipes group usually have a completely black abdomen, we have studied numerous specimens from southern Ukraine, Armenia, and Russian Caucasus that have a mostly black to completely yellow abdomen (even the thorax is partly yellow in some specimens). Because length of postocellar area, sculpture of mesopostnotum, and coloration of tegula, pronotum, and labrum varies among these specimens, it is not clear if they can be associated with any of the species from Northern Europe (lancets are not different from those on Figs 219–220). Recently, Macek (2016) reared three species of this group: P. rufipes from Aquilegia and two species from Thalictrum. The species identified by Macek (2016) as P. sareptana Kuznetzov-Ugamskij, 1924 could actually be P. brevis or P. thalictrivora (if these are different species) based on the pictures given for adults (we have seen the types of P. sareptana and P. similis Kuznetzov-Ugamskij, 1924, which are most similar to P. brevis and could indeed be synonyms of this species). Additional studies are needed to resolve species boundaries (if there are any) of the Thalictrum-feeding complex, and to associate males with females.

If a well defined morphospecies corresponds to exclusive genetic clusters (not mixed with other morphospecies) based on more than one unlinked genetic marker (independently segregating gene regions that reside on different chromosomes or far from each other on the same chromosome) and individuals of that morphospecies have different relationships based on those different markers, then we are probably dealing with a single and well separated species. Unfortunately, such extensive genetic data (many markers and individuals) are not available for most species (due to lack of suitable material or/and high costs associated with sequencing). Nevertheless, it is possible to get some idea about species boundaries based on much smaller amounts of genetic data. Based on our results (see Discussion), within species divergence of mitochondrial genes seems to remain within 5% and based on nuclear genes within 1%. Although divergence higher than 2% in COI barcode sequences is commonly regarded as good evidence for different species (Ratnasingham and Hebert 2013), our data suggest that it is not unusual for within species divergence to be more than 3% or 4%. An example of this is P. albitibia, which can have COI barcode divergence of 4.6%, although neither morphology nor nuclear genes indicate that more than one species is involved. Therefore we do not consider COI divergence of less than 5% as good evidence for the existence of different species if there is no support from morphological and / or nuclear data.

On the other hand, where we have consistent morphological (based on adults or larvae) or ecological (different hosts) evidence without clear genetic evidence (even if based on all three genes we sampled) for the existence of different species within groups of closely related species, we have not synonymised these species. In our view, the genetic evidence we have so far is not sufficient (too few specimens and gene regions sampled) to decide this.

For females, the shape of valvula 3 (apical sawsheath) is an important and relatively stable character for identification, but nevertheless it was necessary to key many species more than once, because of variability or intermediate character states. Still, it might be difficult to key out some species or individiuals, especially when valvula 3 is distorted (which usually happens with specimens dried from alcohol). In these cases lancets should be examined, although they can sometimes be very similar, even among distantly related species.

Males in Pristiphora usually lack good external non-colour characters, and therefore penis valves should be studied in most cases, which are stable within species and often show good differences between the species.

Coloration is the easiest character to observe, but unfortunately it often varies quite a lot within species. The abdomen can be completely pale or (almost) completely black in some species. Thorax coloration varies somewhat less, but can occasionally still vary from completely black to extensively pale. Head coloration is the most stable, in most species being black or with small pale spots. In a minority of species the head tends to be extensively pale around the eyes. Leg coloration can also vary extensively within species, but can often still be used for identification (see the Key). Coloration of the pterostigma is often a useful character for species identification, but it can be problematic in older pinned specimens. In species that normally have a dark pterostigma, the dark coloration can fade, causing the pterostigma to appear pale.

Unreliable characters. We have not used the length of the inner spur of the metatibia (in relation to the length of metatarsus for example) in species identification (e.g. Benson 1958; Zhelochovtsev and Zinovjev 1988, Haris 2006b) because the differences between the species (if there are any) are too small to be easily applied. Shape of the posterior margin of sternum 9 can be quite misleading. Several examined specimens of different species have a distinct notch, compared to other conspecific specimens with the margin entire, but because a notch is very sporadically present and does not correlate with other characters (which might indicate the existence of additional species), it seems to have no value in species identification and was therefore excluded from the key. However, in one exceptional case, the notch is large and deep and even has an outgrowth at the bottom of the notch (Fig. 52). Because it is a single specimen (based on penis valves, it is P. biscalis), it is not clear if this too is an aberration or there is an additional species involved. Vein 2r-rs of the forewing (see Fig. 1 in Prous et al. 2014) is another character that sporadically appears in some specimens, but without any phylogenetic correlation (we have observed it in P. malaisei, P. robusta [http://id.luomus.fi/GL.5198], and P. staudingeri), making it difficult to even recognise these specimens as Pristiphora.

The key for females relies heavily on sawsheaths and ovipositors, and for males on penis valves, as these are most reliable for species identification. For females it is advisable to dissect the saw of one or more specimens of a series and for males it is in most cases necessary to pull out the genital capsule to see at least the tip of the penis valve. Fresh and clean specimens greatly help in species identification: particularly in females, where the shape of sawsheath can be of critical importance. Generally we recommend using the electronic key, which can be significantly faster (when using ‘Best’ and ‘Next Best’ options), easier to use, and more reliable because of the possibility to check more characters.

Females of P. dissimilis are unknown.

Males of P. aphantoneura, P. astragali, P. depressa, and P. subbifida are unknown.