The genera of Nematinae (Hymenoptera, Tenthredinidae)

Recent phylogenetic studies on Nematinae based on DNA sequences have shown extensive incongruencies with current nomenclature of genus-group taxa. Here, we expand previous DNA sequence datasets bas ...


Introduction
The Nematinae is the second-largest subfamily within the Tenthredinidae (Taeger et al. 2010). With about 1250 species, it is surpassed only by the Tenthredininae (about 1700 species) (Taeger and Blank 2011). In the Electronic World Catalogue of Symphyta (ECat-Sym) (Taeger and Blank 2011), 48 genera of Nematinae are currently recognised, but there is still no wider consensus on how many genera should be recognised, and how these should be delimited. Also lacking are comprehensive keys for identification of the genera. Only some regional keys are available (e.g. Benson 1958;Zhelochovtsev and Zinovjev 1988;Goulet 1992) which have to be combined with numerous other publications (Togashi 1964;Taeger 1989;Zinovjev 1993;Wei 1998a;b;Wei and Nie 1998;Zinovjev 2000;Lacourt 2006;Wei and Nie 2008) to cover most of the genera of the world. Many (sub)genera are based on few morphological characters. Furthermore, some of the character states lack discrete differences, making recognition of genera (and therefore species) difficult. Recent phylogenetic analyses based on DNA sequence data (Nyman et al. 2006;2010) have shown that many genera (especially within the so-called 'higher Nematinae') are not monophyletic, indicating a need for a taxonomic revision. Here we expand these previous phylogenetic analyses (based on at most three genes: CoI, Cytb, and EF-1α) with the addition of one nuclear gene (NaK) and more genera to provide a new generic classification of Nematinae. The new classification seems to offer the best prospect of promoting future stability of nomenclature. An extensively illustrated key to genera is also provided. This can be considered as a starting point for the revision of roughly half of the ca. 600 West Palaearctic nematine species, a project funded by the Swedish Taxonomy Initiative (STI). Unfortunately, the key does not solve all the problems in identifying genera. Perhaps the biggest drawback is our current inability to unequivocally separate Euura from Nematus based on morphology. The forthcoming keys to species of individual genera should remove some such remaining ambiguities in identifying genera.

Phylogenetic analyses
We extracted DNA from larval and adult samples stored in 99.5% ethanol at -20 °C by using the DNeasy Tissue Kit (Qiagen, Valencia, CA). Sequence data were collected from the mitochondrial genes Cytochrome oxidase I (CoI; 810 bp) and Cytochrome b (Cytb; 718 bp), and the nuclear genes Elongation factor-1α (EF-1α; two exons of the F2 copy; 777 bp in all) and Sodium-potassium adenosine triphosphatase (NaK; 997 bp). PCR amplification and sequencing of CoI, Cytb, EF-1α, and NaK were done as described earlier (Nyman et al. 2000;Nyman et al. 2006;Leppänen et al. 2012). New sequences have been deposited in GenBank under accession numbers KJ434795-KJ434930. CoI and Cytb sequences of Monocellicampa pruni were extracted from a partial mitochondrial genome available in GenBank (JX566509; Wei et al. 2013b). In some cases, unpublished barcode sequences (CoI; 658 bp) from the BOLD database (http://www.boldsystems.org/) were also included in the analysis to maximize representation of type species of genera.
We constructed two four-gene alignments. The first contains 134 specimens and 3302 base pairs with little missing data. Most specimens in this dataset have sequences from all four genes, but 19 specimens are missing one gene and 7 specimens are missing two genes. In order to examine the relationships between type species of the genus-group names, a second dataset of 79 specimens and 3537 bp was constructed. This alignment is longer because of the addition of the 658-bp barcode region of CoI (Hebert et al. 2003): 423 bp of this region overlaps with the 5' end of the 810-bp CoI portion used in the first dataset. When the overlapping region was identical between two different conspecific specimens, a single composite terminal taxon was created to minimize the missing cells in the dataset. In this second dataset, 21 specimens are missing one gene, 8 specimens are missing two genes, and 15 specimens are missing three genes. Four non-type species, Pristicampus incisus (Lindqvist, 1970), Paranematus tulunensis (Vikberg, 1972), Craterocercus fraternalis (Norton, 1872), and Susana annulata D.R. Smith, 1969, were added to the dataset, because the amount of sequence data for the respective type species (Pristicampus arcticus (Lindqvist, 1959) [422 bp of the CoI barcode region], Paranematus wahlbergi (Thomson, 1871) [658 bp of CoI], Craterocercus obtusus (Klug, 1816) [658 bp of CoI], and Susana cupressi Rohwer and Middleton, 1932 [997 bp of NaK]) was insufficient to reliably estimate their phylogenetic position.
When we exclude genus names which are based on fossils, or are taxonomically unplaced, there are 78 genera based on different type species. Of these, 62 type species have DNA data. Of the remaining 16 type species, most can be associated using morphology with species for which DNA data is available, but Anhoplocampa, Armenocampus, Dinematus, Katsujia, Megadineura, Nescianeura and Renonerva still lack DNA data.
We performed Bayesian phylogenetic analyses in MrBayes v. 3.2.2 (Ronquist et al. 2012) and maximum likelihood (ML) analyses in RAxML v. 7.6.6 (Stamatakis 2006;Stamatakis et al. 2008) using the CIPRES Science Gateway (Miller et al. 2010) at http://www.phylo.org/index.php/portal/. The dataset was partitioned by genes, and the best-fitting DNA substitution model for each gene was selected using jModel-Test 2.1.4 (Darriba et al. 2012), which uses PhyML (Guindon and Gascuel 2003) for likelihood calculations. Model selection was done using five substitution schemes (including parameters for base frequencies, gamma-distributed rate variation across sites (G) with four categories, and a proportion of invariable sites (I), altogether 40 different models) on the basis of the Akaike Information Criterion (AIC). The best-fitting model for all four genes was GTR+I+G, which was used in MrBayes. In the RAxML runs, proportion of invariable sites (I) was not used as recommended in the manual of the program. In MrBayes, we used default priors, and each of the four partitions was allowed to have its own unlinked substitution model. We ran two parallel runs having four incrementally heated chains for 20 million generations, while sampling trees from the current cold chain every 1000 generations. We discarded 5000 trees sampled prior to reaching chain stationarity as a burn-in from both runs, and the remaining 15001 trees were used to calculate a 50% majority consensus rule tree, showing all groupings with posterior probability more than 0.5. In RAxML a separate GTR+G model was employed for each gene, and node supports were evaluated based on 500 bootstrap replicates. The root of the phylogenetic trees on Figs 2-5 was placed between Nematinae and other representatives of Tenthredinidae in the dataset. For the subfamilial phylogenetic relationships and placement of Nematinae within Tenthredinidae, see Malm and Nyman (2014).

Phylogenetic analyses
Our focus here is nomenclature and delimitation of genera. Accordingly, we discuss the implications of phylogenetic results only from this perspective. The results largely agree with previous analyses (Nyman et al. 2006;2010) regarding congruence with current taxonomy: Hemichroa militaris is distinct from other Hemichroa; Stauronematus does not belong to Pristiphora; Pristiphora, Nematus, and some other genera (as delimited in Taeger et al. 2010) in the Euura clade are non-monophyletic; and most species belong to two large clades, Euura and Pristiphora . (Note that Dineura virididorsata in Nyman et al. 2006 was a misidentified larva of Nematinus acuminatus; the name has been corrected in GenBank, accessions DQ302173 and DQ302261). The only important difference when it comes to delimitation of genera compared to the previous Figure 2. Phylogenetic tree of Nematinae based on the Bayesian analysis. Numbers above or right of branches show Bayesian posterior probabilities (PP) followed by bootstrap proportions (%, BP) from the corresponding ML analysis. Branches receiving maximum support (PP=1, BP=100%) are denoted by a black dot. Support values for weakly supported branches (PP<0.9 and/or BP<70) are not shown. Euura and Pristiphora clades are collapsed here, but are fully shown in Figs 3 and 4. The inset shows the outline of the full tree including the Euura and Pristiphora clades. Nomenclature is according to Taeger et al. (2010). Voucher ID numbers (e.g. G6) correspond to specimen identifiers in previous phylogenetic trees with different species names. The scale bar shows the number of estimated substitutions per nucleotide position.  Fig. 2. Numbers above or right of branches show Bayesian posterior probabilities (PP) followed by bootstrap proportions (%, BP) from the corresponding ML analysis. Branches receiving maximum support (PP=1, BP=100%) are denoted by a black dot. Support values for weakly supported branches (PP<0.9 and/or BP<70) are not shown. The inset shows the outline of the full tree. Nomenclature is according to Taeger et al. (2010). Voucher ID numbers (e.g. 6b) correspond to specimen identifiers in previous phylogenetic trees with different species names. The scale bar shows the number of estimated substitutions per nucleotide position.  Fig. 2. Numbers above or to the right of branches show Bayesian posterior probabilities (PP) followed by bootstrap proportions (%, BP) from the corresponding ML analysis. Branches receiving maximum support (PP=1, BP=100%) are denoted by a black dot. Support values for weakly supported branches (PP<0.9 and/or BP<70) are not shown. The inset shows the outline of the full tree. Nomenclature is according to Taeger et al. (2010). Voucher ID numbers (e.g. F2) correspond to specimen identifiers in previous phylogenetic trees with different species names. The scale bar shows the number of estimated substitutions per nucleotide position.
analyses is the phylogenetic position of the type species of Nematus, N. lucidus. The addition of the nuclear NaK gene to the dataset causes N. lucidus, as well as Mesoneura, Fagineura, Craesus, and some other species of Nematus to move away from the Euura clade (placed closer in Nyman et al. 2006;2010) and approach the Pristiphora clade (Figs 2 and 5). Considering previous phylogenetic results and the results reported here (Figs 2-5), we propose in the Taxonomy section several changes compared to Taeger et al. (2010). Advantages and disadvantages of this and alternative taxonomies are highlighted in the Discussion section. Figure 5. Phylogenetic relationships of type species for which at least 400 bp of the CoI barcode region was available. The tree was reconstructed according to a ML analysis allowing a separate GTR+I+G4 model of substitution for each gene. Numbers above or right of branches show bootstrap proportions (%, BP) followed by posterior probabilities (PP) from the corresponding Bayesian analysis. Branches receiving maximum support (BP=100%, PP=1) are denoted by a black dot. Support values for weakly supported branches (PP<0.9 and BP<50) are not shown. Nematinae was not monophyletic in the Bayesian analysis, because Moricella rufonota (only 658 bp of CoI available) was weakly placed as sister to the Strongylogaster+Nesoselan dria clade. Some terminals in the dataset are composites of two conspecific specimens (indicated by two ID numbers, for example J3 and DEIGISHym11927 or F2 and BC ZSM HYM 09367). Four non-type species (Pristicampus incisus, Paranematus tulunensis, Craterocercus fraternalis, and Susana annulata) were also included in the analysis, because the CoI or NaK (for S. cupressi) sequences available for the type species of the respective genera were not sufficient to place them reliably in the tree. Type species in bold are the basis for genera defined here. Outgroup taxa have been collapsed for simplicity. Voucher ID numbers (e.g. F2) correspond to specimen identifiers in previous phylogenetic trees with different species names. The scale bar shows the number of estimated substitutions per nucleotide position.

Key
For some smaller genera and species groups there are comprehensive recent keys to species available, which we have referred to in the key. Because the morphological separation of Euura, Pristiphora, and Nematus can be difficult (see Discussion), there are two places (couplets 15 and 17) in the key where we have not separated them completely. Instead, we have given short descriptions of a few minor groups which run together with the genus intended to be keyed out. Full resolution of these problems requires species-level revisions. The key is arranged in alternating pages of text and plates. The couplets are illustrated by a plate on the facing page. We recommend using the key with two pages side by side, so that couplets and corresponding figures are simultaneously visible.

11(10) a
Height of eye in lateral view about 2-3 times as long as distance from dorsal margin of eye to dorsalmost point of head. b Sawsheath emarginate in dorsal view. c Tangium of lancet with campaniform sensilla ("pores"). d Valviceps not divided into pseudoceps and paravalva, without valvispina.
Small, 4.0-5.5 mm, black and yellow-brown; thorax (except pronotum), abdomen above, head (almost entirely) and antennae black; clypeus, labrum and base of mandibles white. In the forewing costa same colour as pterostigma (usually pale). Three species of the P. arctica group. Formerly Pristicampus. Revision by Zinovjev (1993) About four species of the N. abbotii group, including N. princeps Zaddach, 1876. Key to Nearctic species by Smith (2008 Body 4.5-7.0 mm, black or with yellowish or reddish abdomen, sometimes also thorax largely reddish; stigma dark brown; hind tibia slightly widened with an indistinct longitudinal groove; metatarsomere 1 cylindrical, without a groove; sawsheath short and rounded in lateral view, wide in dorsal view; cerci longer than sheath; lancet with lateral spines on annuli; male penis valve is slightly curved, with stout valvispina, and a rather low, rounded adjacent lobe. About six species of the N. erythrogaster group, including N. lucens (Enslin, 1918) and N. umbratus Thomson, 1871. Keys to species by Liston et al. (2006), Smith Small, 4.0-5.5 mm, black and yellowish brown; thorax (except pronotum), abdomen above, head (almost entirely) and antennae black; clypeus, labrum and base of mandibles white. Forewing costa same colour as pterostigma (usually pale). Three species of the P. arctica group. Formerly Pristicampus. Revision by Zinovjev (1993 Rohwer, 1911 -aa Abdominal tergum IX in lateral view in female less than 2 times as long as tergum VIII. bb Penis valve in male without filament. cc Anterior depressed section of metepisternum along metepimeroepisternal suture 0.1-0.3 times as wide as posterior section. dd Cell 1Rs and 2Rs often fused because vein 2r-m absent (see Fig. 9a

34(33) a Clypeus widely and deeply emarginate.
b Malar space clearly shorter than diameter of front ocellus. c Anterior half of mesepimeron partly or completely covered with setae. d Vein 2r-rs of fore wing present (see Fig. 5a). e Nearctic.
Six species. Key to most of the species by Smith (1969a Wong, 1977 37(36) a Anterior protibial spur without velum, but with hairs.

49(48) a Cell A of hind wing rectangular, widening at apex with vein 1A
beginning from anterior margin of cell A. b Subapical tooth of tarsal claw long and slender.
Two species. Key to species by D.R.   aa Cell A of hind wing not rectangular, tapering at apex with vein 1A beginning approximately from centre of cell A. bb Subapical tooth of tarsal claw absent.

Secondary homonymy of species names
The new and much wider circumscription of Euura adopted in this work involves the synonymy of several partly species-rich nominal genera. As a result, a number of species names become secondary homonyms when they are placed in Euura. In all except five of the 25 cases listed below, the senior homonym has been applied after 1899 to a taxon considered to be valid. Article 23.9.1 of the International Code of Zoological Nomenclature (ICZN 1999) is therefore not fulfilled in these cases, and the junior homonyms require replacement. In the five remaining cases (Amauronematus poppii, Euura cinereae, Pontopristia punctulata, Pteronidea brachycera and P. curticornis) the junior homonym has not been used for a particular taxon, as its presumed valid name, in at least 25 works, published by at least 10 authors in the immediately preceding 50 years and encompassing a span of not less than 10 years. The second condition stipulated by Article 23.9.1 is thus not met, and these homonyms also require replacement. Only dealt with below are cases of secondary homonymy where the taxonomy of both species at present seems reasonably clear. In several remaining cases, the validity of the species denoted by the junior homonym is highly questionable and it is therefore not desirable to replace them now. The decision on whether a replacement name is necessary should be made once the relevant groups have been better studied. The replacement names proposed are suggested by the authors of the present paper who studied the individual cases, except for the species described in Euura, Phyllocolpa and Pontania. Jens-Peter Kopelke allowed us to publish, in his own words, the relevant replacement names for secondary homonyms of the latter.

Discussion
It has been evident since 2006 (Nyman et al. 2006) that changes are necessary in the classification of the Nematinae to reflect the current understanding of their phylogeny. Most problematic are the 'higher' Nematinae as defined in Nyman et al. (2006). Subsequent analyses (Nyman et al. 2010; this work), including more taxa and genes, have confirmed the existence of two well-supported and species-rich clades within the 'higher' Nematinae. Although there are several ways to divide the phylogenetic tree into genera, we decided to treat these two major clades as Pristiphora and Euura, which are the oldest available genus group names for these clades. Although it might seem preferable to use the name Nematus instead of Euura, because most of the species have been placed in Nematus in the past (e.g. Zhelochovtsev and Zinovjev 1988), the phylogenetic placement of the type species Nematus lucidus is unfortunately not stable. Previous analyses (Nyman et al. 2006;2010) placed N. lucidus closer to Euura, but this relationship is no longer supported in our analysis that also includes NaK sequences. Even if later analyses with more markers support a N. lucidus-Euura clade, and Euura could be subsumed under Nematus, the possibility still remains to treat these potential sister groups (it is unlikely that N. lucidus falls within Euura) as separate genera (Fig. 6). Regardless of the position of N. lucidus, Euura is a well-supported clade and therefore is a good candidate for a genus. The deeper relationships within this clade, however, are genetically poorly resolved and often at odds with current nomenclature, which argues against creating or maintaining genera within this clade. Within the Euura clade, another possibility would be to retain monophyletic genera, modify those which are not monophyletic, and create new ones for the remaining species. However, this option would result in the proliferation of morphologically and genetically poorly defined, unstable genera (many of which would be monotypic) because of weakly established phylogenetic relationships. Attempting to maintain the status quo is equally unlikely to result in nomenclatural stability, because the creation of new genera and the redefinition of established ones would be expected to continue in the foreseeable future. It would also be possible to include Pristiphora, Mesoneura, Fagineura, and Euura in Nematus (this clade is well supported), but this would result in many more nomenclatural changes (new combinations and secondary homonyms) compared to the "Euura" option.
In the other big clade, Pristiphora, which is here treated as a genus, the nomenclatural changes are fortunately less extensive than in Euura. Splitting up Pristiphora in order to maintain a few small genera, such as Melastola, Pristola, Nepionema, Neopareophora and Pristicampus, would cause problems similar to those that would result from splitting up Euura. The creation of several poorly defined genera with weakly resolved phylogenetic relationships would certainly lead to nomenclatural instability in the future.
We want to emphasize that although there are uncertainties regarding many relationships in the phylogeny of Nematinae (including the exact placement of N. lucidus), this does not mean that all or even most relationships are unreliably reconstructed. The support for the two largest clades, Pristiphora and Euura, has increased with the addition of NaK (compared to Nyman et al. 2010, ML bootstrap support went up from 98% to 100% for Euura and from 89% to 99% for Pristiphora).
Some taxa in Euura, Nematus, and Pristiphora are monophyletic and relatively well defined (Euura s.s., a large part of Amauronematus, Craesus, Pristicampus etc.), which would allow recognition of these groups. However, for the sake of nomenclatural stability we strongly recommend that the formal acceptance of these clades as genera should be avoided. Preferably such infrageneric groups should be referred to as species groups (within the context of ICZN) or defined according to the PhyloCode (Cantino and de Queiroz 2010). Figure 6. Three possible relationships between the Euura, Pristiphora, and Nematus clades. All are consistent with treatment of these three clades as separate genera, or combining all of them as a single genus Nematus. Including Euura, but not Pristiphora, within Nematus is consistent only with the tree shown in the middle (supported by previous analyses: Nyman et al. 2006;2010).
Although the Nematus clade received less support than Euura and Pristiphora, we did not attempt to divide it further because it includes only few species and most of them already have the generic name Nematus. The only exception is Craesus as it is usually considered to represent a separate genus. However, it is nearly indistinguishable from the erythrogaster-group of Nematus when we exclude the exceptionally expanded hind tibia and metatarsomere 1 of Craesus (Smith 2008). If future phylogenetic research shows that this group is not monophyletic, Nematus can, if necessary, still be divided into smaller genera. However, this will affect only a small number of species.
The rather broad definitions of genera like Pristiphora and Euura, and the resulting nomenclatural changes will certainly cause controversy, but in our opinion this is the best solution in light of current evidence. Pristiphora and Euura are strongly supported clades (and it is unlikely that future data will challenge this), but their relationships to other 'higher' Nematinae, as well as basal branching patterns within Pristiphora and Euura, remain controversial. We wish to stabilize the nomenclature of 'higher' Nematinae by considering these clades as genera, so that in the future it is possible to concentrate on actually studying these sawflies without having to deal with constant name changes. The downside of our approach is that we have to make many nomenclatural changes, especially concerning Euura, but when accepted, it will be a once-only event.
The other change, compared to the World Catalog (Taeger et al. 2010), is the transfer of Nematus militaris Cresson, 1880 to Dineura, creating Dineura militaris (Cresson, 1880), comb. n. (previously in Hemichroa). According to DNA sequence data, Dineura militaris forms a clade with Dineura s.s. to the exclusion of other genera in Dineurini (Figs 2 and 5). Absence of a velum in Dineura militaris and other Dineura might be a synapomorphy for these taxa.
Based on current data, the higher level relationships (i.e. between genera) within Nematinae are generally not well supported, but there are few clades worth mentioning: Dineurini, Pseudodineurini and 'higher' Nematinae (Fig. 2). Of these three clades, composition of only Pseudodineurini matches with the morphology based classifications (e.g. Zhelochovtsev and Zinovjev 1988;Goulet 1992). Composition of Dineurini and 'higher' Nematinae is the same as found by Nyman et al. (2006). However, before any tribal classification of Nematinae is proposed, molecularly yet unsampled genera and more sequence data should be gathered to confidently resolve the phylogeny of genera.
Although we are far from having resolved all problems involved in the definition and identification of the genera of Nematinae, we are confident that our pictorial approach using photographs of actual specimens rather than drawings makes identification much easier than before in this challenging group. We regard the key as a starting point for future improvements: the main remaining shortcomings involve Euura, Nematus and Pristiphora, two of which (Euura and Pristiphora) comprise over 75% of nematine species. Nevertheless, most species of these genera should key out correctly, as only a few Pristiphora species are likely to run to Euura or vice versa. Pristiphora has campaniform sensilla on the tangium of the lancet and generally a more or less truncate clypeus and a swollen apex of vein C, while Euura and Nematus lack campaniform sensilla and generally have an emarginate clypeus and a less swollen apex of vein C. While some groups of Nematus are differentiated morphologically from Euura by characters of metatarsomere 1 (see couplets 12-13 in the Key), we are currently unaware of unequivocal characters to distinguish the Nematus wahlbergi and erythrogaster groups from Euura. In addition, there are currently many, mostly Nearctic species that are placed in Nematus or Pteronidea in Taeger et al. (2010), but which cannot be associated confidently with our current concepts of Nematus or Euura due to lack of recent revisions and/or DNA sequence data. Solving these problems requires species-level revisions, regardless of how genera are defined. For Euura, Nematus and Pristiphora, a key to species level is urgently needed, rather than having an intermediate step keying out genera or subgenera. Work is currently under way in this direction for the West Palaearctic species.