DNA barcoding of Palaearctic Ulidiidae (Diptera: Tephritoidea): morphology, DNA evolution, and Markov codon models

Here, for the first time, we report a barcoding survey of the dipterian family Ulidiidae (with two subfamilies Ulidiinae and Otitinae) coupled with morphology. To date, this is the first comprehensive analysis of the family that relies on molecular data. To reconstruct probable higher-level phylogenetic relationships between the genera of Ulidiidae, we exploited maximum likelihood and Bayesian inference, and additionally, we utilized a modern Markov model of codon substitutions for protein-coding genes coupled with the maximum likelihood approach to establish more realistic evolutionary scenarios connecting Ulidiinae and Otitinae. Though we found morphological synapomorphic characters that clearly distinguish two groups of genera, formerly relating to two subfamilies, the monophyly of the clade Otitinae was not firmly supported by molecular phylogenetic reconstructions. The subfamily Ulidiinae was recovered as the basal likely paraphyletic group with high reliability. Overall, our results suggest the inclusion of the genera Homalocephala and Seioptera into the tribe Seiopterini (Otitinae). Three genera of Ulidiinae (Physiphora, Timia, Ulidia), eight genera of Otitinae (Ceroxys, Herina, Melieria, Myennis, Otites, Pseudotephritis, Seioptera, Tetanops), and the genus of Homalocephala with still unconfirmed systematic status were included in the analysis. On all phylogenetic trees obtained in our survey, there is a clear tendency of clustering at the genus level with separation of (Physiphora + (Timia + Ulidia)) (Ulidiinae) and (Otitinae + Homalocephala). The genus Homalocephala takes basal or subbasal position relatively to Otitinae. Phylogenetic reconstruction based on Markov models of codon evolution provided a good resolution for our limited dataset.


Background
The true fruit fly family Ulidiidae Macquart, 1835 consists of subfamilies Ulidiinae Macquart, 1835 and Otitinae Westwood, 1840 and belongs to the superfamily Tephritoidea of the acalyptrate Diptera. Inability to separate Ulidiidae from other Tephritoidea families using morphological characters alone introduced great uncertainty about its taxonomic status including multiple divisions versus unifications into a single family (Figure 1). The subfamily Ulidiinae includes approximately 430 species assigned to about 60 genera. There have been 91 species of Ulidiinae described from 3 genera in the Palaearctic. These flies are small-to medium-sized (2 to 13 mm). Ulidiins consist of diversely colored flies (black or reddish-yellow, often dull gray, sometimes with a metallic green or blue shine) with predominantly hyaline wings, often with sub-basal and apical dark spots. This subfamily occurs nearly worldwide, with more than half of the species and 75% of the genera in the Neotropical region (Kameneva 2008). About 100 species of Ulidiinae belonging to 5 genera inhabit the Palaearctic territory. The subfamily Otitinae includes about 260 species assigned to 31 genera, distributed mainly in the Palaearctic region and the Americas. There have been 128 species of Otitinae described from 12 genera in the Palaearctic. Otitins are generally small-to medium-sized (3 to 12 mm) flies with diverse coloration; their wings are either predominantly hyaline or with a number of dark spots. Both subfamilies are characterized by a high level of species diversification and by a varying degree of morphological genus differentiation. Within the superfamily Tephritoidea, the family Ulidiidae occupy a proximal position to the root of the phylogenetic tree ( Figure 2) and thus play a crucial role in understanding the major evolutionary trajectories of tephritoidous flies. To date, researches have not reached any strict consensus regarding the taxonomic status of the group Ulidiinae + Otitinae.
Loew (1868) was the first to propose a detailed classification of the family named Ortalidae (Figure 1). Within the family boundaries, he discerned seven groups: Pyrgotina, Ortalina, Cephalina, Platistoma, Pterocallina, Ulidiina, and Richardiina. Later, Ulidiinae, Pterocallinae, and Ortalinae have often been considered as the subfamilies in Muscaridae (Hendel 1910a, b) or as the separate families Ulidiidae, Pterocallidae (syn. Ulidiidae), and Ortalidae (syn. Otitidae) (Hendel 1928;Stackelberg 1933). Hennig (1939Hennig ( , 1940Hennig ( , 1973a discriminated Otitidae (including true Pterocallinae) and Ulidiidae (including Euxestinae) as independent families. Hennig (1939Hennig ( , 1940 showed that these two families might be only differentiated by the distiphallus structure and by the presence/absence of setulae on vein R 1 . Solid synapomorphic characters uniting Ulidiidae and Otitidae into a single monophyletic group were not found. In addition, Hennig (1939Hennig ( , 1940 included the genus Homalocephala Zetterstedt, 1838 into the family Ulidiidae. Nonetheless, Homalocephala was later moved to the family Otitidae (Hennig 1973a, b), whereas Homalocephala had been previously included either in Otitidae (Hendel 1910a, b) or in Pterocallidae (Stackelberg 1933). Most European dipterists (Richter 1970a, b;Soós 1984;Zaitzev 1984;Krivosheina and Zaitzev 1989;Greve 1998;Nartshuk 2003) use two separate family names, Ulidiidae and Otitidae, following Hennig's earlier opinion (Hennig 1939(Hennig , 1940(Hennig , 1973a. Based on American fly collections, Curran (1934aCurran ( , b, 1944 proposed an alternative classification combining Ulidiinae with Otitinae into a single family called Otitidae. Besides Ulidiidae and Otitidae, he also included Pterocallinae, Richardiinae, and Platystomatinae in this family. At this time, the concept of synapomorphic and symplesiomorphic similarities was not yet developed; therefore, observed symplesiomorphic characters had been viewed as the evidence of systematic relationships. Thus, Ulidiinae + Otitinae + Pterocallinae + Richardiinae + Platystomatinae was thought to be a natural group, and the problem of acceptance of either five families or one family with five subfamilies had mostly classification agreement reasons. Steyskal (1961Steyskal ( , 1965Steyskal ( , 1968Steyskal ( , 1977Steyskal ( , 1987 modified existing classification by the exclusion of Richardiidae and Platystomatidae, treating them as separate families, that was also supported by McAlpine (1977McAlpine ( , 1989. Rohdendorf (1974Rohdendorf ( , 1977, in his system of Diptera, considered the single family Otitidae as including the following four subfamilies: Otitinae, Ulidiinae, Euxestinae, and Pterocallinae. In addition, Evenhuis (1989) determined a single subfamily Lipsaniinae for genus Lipsana Enderlein, 1940 (current syn. Physiphora of the family Otitidae) from the Australian region. Kameneva and Korneyev (1994) showed that the genera Seioptera Kirby, 1817 and Pseudoseioptera Stackelberg, 1955 (Otitinae) are phylogenetically close to Homalocephala (Ulidiinae) and combined them into the tribe Seiopterini (Ulidiidae: Ulidiinae). The monophyly of the tribe is supported by several morphological characters: the absence of anepisternal bristles and the presence of two pairs of katepisternal bristles, the epandrium expanded in the anterocaudal direction, the distiphallus pubescent or short spinulose in its medial part, no sclerotized preglans or long bristles, and two pairs of spermathecae in females. Therefore, Korneyev, 1994, 2006;Korneyev, 1999) included the rest of the Palaearctic Otitidae into the tribes Cephaliini, Otitini, and Myennidini, whereas the remaining Ulidiidae (without genus Homalocephala) were included into the tribes Ulidiini and Pterocallini, while considering all of the tribes within the scope of Ulidiidae. For uniting Ulidiinae and Otitinae into a single family, Kameneva and Korneyev used the following characters: non-developed fronto-orbital plates without strong setae (except in Chaetopsis Loew and close genera); absence of presutural supra-alar (except in Dyscrasis Hendel and some species of Otites Latreille) and katepimeral setae (presutural supra-alar present in the ground plan of most other families; katepimeral setae or setulae present in Platystomatidae, Ctenostylidae, Pyrgotidae, and Tephritidae); slightly bowed continuous Sc vein, neither bent anteriorly at right angle nor constricted before apex (bent anteriorly at right angle in Pyrgotidae Toxurini and most Tephritidae; constricted or broken before apex in Ctenostylidae, most Tephritidae, and some Pyrgotidae); R 1 vein (See figure on previous page.) Figure 1 Historically most significant taxonomic changes and modifications of Ulidiidae at the family level. Circles represent families or subfamilies. Dashed line corresponds to the position of Homalocephala throughout the development of the system. Dot-dashed line corresponds to the position of Seioptera throughout the development of the system. Rectangles combine subfamilies or groups into families, where gray color denotes major classification trends. The arrow indicates inclusion of American species of Otitidae by Curran (1934a, b) into Otitinae of the Ulidiinae + Otitinae group.
usually setulose only on apical portion (bare in all Ulidiini and most Lipsanini, and certain species of other tribes sometimes either bare or entirely setulose; bare in most or all Piophilidae, Pallopteridae, Lonchaeidae, and Richardiidae, and entirely setulose in Platystomatidae, Ctenostylidae, Pyrgotidae, and Tephritidae); a basal cubital (bcu) cell with triangular posterodistal lobe, except in Homalocephala Zetterstedt and few Pterocallini (absent in lower Tephritoidea, all Platystomatidae and Ctenostylidae, and in few genera of Pyrgotidae and Tephritidae); absence of male tergite 6 (rudimentary in few Lonchaeidae and Pallopteridae); bare male sternite 6 (setulose in lower Tephritoidea); absence of abdominal spiracles 6 and 7 in males (present in many lower Tephritoidea); phallus, when at rest, stored in pouch on ventral aspect of abdomen; apex of the phallus without separated glans (stored on dorsal side underneath tergite 6 with glans in other higher Tephritoidea); bare taeniae of eversible membrane without trichoid sensilla (with trichoid sensilla along whole length in most Piophilidae, Pallopteridae, and Lonchaeidae) and with two ducts of spermathecae; apically bifurcated single duct with two spermathecae, except in Lipsanini (two spermathecae on two ducts) (three spermathecal ducts in ground plan of Pyrgotidae and Tephritidae; two of them bifurcated close to vagina) (Kameneva and Korneyev 2006).
Since this is currently the most thorough and comprehensive taxonomic review of the group, we followed their hypothesis to consider the single family Ulidiidae with two subfamilies Ulidiinae and Otitinae.
Molecular phylogenetic studies of the superfamily Tephritoidea were conducted at the genus, subfamily, and family levels with the use of various molecular markers: 16S rDNA and COI (Zhang et al. 2010), COII (Smith and Bush 1997), 16S rDNA (Han et al. 2006), and 12S, 16S rDNA and COII Ro 2005, 2009). Often phylogenetic inferences reached agreement with morphological data. Nevertheless, in most cases, a reconstruction of  Gibson et al. 2010;Wiegmann et al. 2011). Outgroups were selected based on current hypotheses of phylogeny: the superfamily Nothyboidea (including Psilidae) form a sister group to the superfamily Tephritoidea (including Ulidiidae). Drosophilidae (superfamily Nothyboidea) is a sister group to Nothyboidea + Tephritoidea.
phylogenetic relationships among species/genera was based on relatively small sample surveys, which do not represent the full taxonomic diversity of the superfamily Tephritoidea. In addition, the phylogenetic schemes derived by different authors, with the implementation of various species samples and sets of molecular markers, make it difficult to compare these results. Importantly, the number of species from the family Ulidiidae included in the studies mentioned above was very small (two to three species, one individual).
In our study, we compared the external morphology and genital structures of Ulidiidae coupled with rigorous molecular phylogenetic analyses, in order to clarify its interspecific relationships. For species identification, we used external and genital morphology; for phylogenetic reconstruction, we performed stable morphological characters and sequence analysis based on a commonly used COI mitochondrial molecular marker. We found justification for the uniting of Ulidiinae and Otitinae into a single family.

Taxon sampling and morphological analysis
Sampling for subsequent phylogenetic analysis was focused on Ulidiidae flies of the Palaearctic region (Additional file 1). The specimens were identified, labeled, and stored in the fly collection of Zoological Museum at Lomonosov Moscow State University, Moscow, Russia (ZMUM).
For the investigation of external morphology and the anatomy of terminal male and female structures, in total, approximately 6,800 fly specimens belonging to 82 species of 3 genera of the subfamily Ulidiinae and 2,700 fly specimens from 107 species of 12 genera of the subfamily Otitinae were analyzed from museum collections. Morphological characters were examined and visualized under a stereomicroscope Stemi SV11 Apo Carl Zeiss (Oberkochen, Germany) and microscope Olympus CX41 (Shinjuku, Tokyo, Japan). The male and female terminalia were dissected out, treated with 10% KOH solution at approximately 80°C for several minutes, and observed in a droplet of glycerol under a compound light microscope. Photographic series were created with the use of CombineZM software (Hadley 2007).
This study is based mainly on the material deposited in the Zoological Institute of the Russian Academy of Sciences, Saint-Petersburg, Russia (ZISP), the Zoological Museum at Lomonosov Moscow State University, Moscow, Russia (ZMUM), the Hungarian Natural History Museum, Budapest, Hungary (HNHM), the Museum of Natural History at Humboldt University in Berlin, Germany (ZMHB), and Naturhistorisches Museum Wien, Austria (NHMW).
The substitution model of DNA sequence evolution was chosen among 88 competing models using a model test implemented in the IQ-Tree v. 0.9.4 software (http://www.cibiv.at/software/iqtree) (Minh et al. 2013) according to the model log-likelihood values. The TVM model, with a rate variation among sites following a gamma distribution (G), was selected according to the highest log-likelihood of −7706.4766. This model was used to construct a maximum likelihood (ML) tree. To do this, we used the IQ-Tree v. 0.9.4 software. In order to obtain branch support values, 1,000 bootstrap replicates were performed in the same software utilizing the chosen evolutionary model. In addition, a replicated Bayesian analysis coupled with Markov chain Monte Carlo (MCMC) simulations starting from an initial random tree was performed in MrBayes v 3.2.1. (http:// mrbayes.sourceforge.net/) (Ronquist and Huelsenbeck 2003) for four million generations in two independent runs with a sampling frequency every 500 generations. All trees that were sampled before the two runs reached a value of standard deviation less than 0.015 were discarded as burn-in. The resulting trees were summarized to a consensus, and posterior probabilities (PP) as branch statistical support were then calculated. Since the ML and Bayesian trees agreed on one topology, we drew the bootstrap support values (in %) from 1,000 bootstrap trees onto the Bayesian tree using the IQ-Tree software, thus combining bootstrap support (BS) with posterior probability values.
In addition to a typical phylogenetic analysis based on common models of nucleotide substitutions in DNA sequence, we reconstructed phylogenetic relationships utilizing a model of codon evolution. Markov models of codon substitution have emerged as a powerful method modeling biologically more realistic scenarios for proteincoding sequences (Gil et al. 2013). The most important advantages of codon models over DNA models are being achieved by naturally incorporating selection regimes acting on protein and the structure of the genetic code as relevant model parameters. For further phylogenetic inference, the YAP parametric codon model was a priori selected, reasoning that estimation of the non-synonymous to synonymous substitution ratio (ω) is not affected by base composition at each codon compared to other codon models, namely GY and MG (for details, see Yap et al. (2010)). The selective variability (distribution of ω ratios) among codon sites was modeled using the model M3 with three discrete site classes (Yang et al. 2000). Empirical codon frequencies were estimated under CFx4 option, which enables correction for absence of stop codons (Kosakovsky Pond et al. 2010). Then, ten initial trees generated by the BioNJ method corresponding to ten random starts were estimated. Optimization of each tree topology was carried out using simultaneously two ML heuristic tree search algorithms -nearest neighbor interchange (NNI) and subtree pruning and regrafting (SPR). The most likely tree was obtained with SPR moves (log-likelihood = −6781.1666). The branch support was assessed by the Bayesian-like modification of the approximate likelihood ratio test (aBayes) (Anisimova et al. 2011). aBayes was defined to be the most powerful among other branch support methods based on an approximate likelihood ratio test. Although aBayes supports tend to be very similar to PP supports, the aBayes method exhibits more conservative behavior and robustness to model violations (Anisimova et al. 2011). All of the above procedures and codon ML tree construction were implemented in Codon-PhyML Ver1.0 (http://sourceforge.net/projects/codonphyml/) (Gil et al. 2013).
For the barcoding data, the distance matrix was produced under the Kimura two-parameter (K2P) (Kimura 1980) model of base substitution with uniform rates using MEGA 5 (Tamura et al. 2011). In addition, we calculated uncorrected p-distances as an alternative distance metric that may be preferable to K2P for short, closely related sequences. The intraspecific variation and interspecific divergence levels were investigated by comparing pairwise K2P distances within and between fly species.

Barcoding gap
The K2P model has been recently criticized as being inappropriate in some cases for divergence estimation and having no evident advantages over uncorrected p-distance (Srivathsan and Meier 2012). Indeed, a comparison of K2P with p-distances in our dataset showed high correlation between both distance metrics (Pearson correlation = 0.9980, P < 2.2e − 16; Additional file 4). However, for barcoding gap assessment, we used the K2P distance, which is widely accepted in DNA barcode literature. The interspecific genetic K2P distances averaged at 0.1483 (SD = 0.0262, n = 2356), while conspecific distances averaged at 0.0078 (SD = 0.0154, n = 59). An apparent non-significant (Wilcoxon rank-sum test, W = 100.50, P < 2.2e − 16) overlap was found between intraspecific and interspecific distributions of K2P distances. Interspecific distance values between Homalocephala albitarsis Zetterstedt, 1838 and Homalocephala biumbrata (Wahlberg, 1839) created the overlap with intraspecific K2P distribution (Figure 3, first asterisk). These two species are morphologically close based on the following characters: face, parafacials, and genae are white; katepisternum is silvery pubescent and with long bright setae on its inner margin. However, they are currently considered as independent species (Andersson 1991) based on median wing spot size, apical wing spot color intensity, and a surface structure of scutellum. Therefore, the low value of K2P distance (0.0109) occurs most likely due to a recent split between these two species. Due to the high intraspecific sequence variation between Melieria cana (Loew, 1858), two K2P distances (Figure 3, second asterisk) formed an overlap with the overall K2P divergence distribution. Interestingly, left-tail distances located on the 0.03 to 0.08 interval of the interspecific K2P distribution were generated mostly by pairwise comparisons between species within the Melieria R.-D., 1830 genus. This observation may indicate that species within the genus Melieria are currently undergoing speciation events. To rule out the possibility that K2P distances may be affected by sequence length parameter, we compared the distribution of K2P within short (<400 bp) and long (>400 bp) sequences. No significant differences were observed (Wilcoxon rank-sum test, W = 713,950.50, P = 0.06; Additional file 5).

Phylogenetic analysis
To test whether our sequence dataset possesses the phylogenetic signal necessary for tree reconstruction, we performed Xia's substitution saturation method based on the concept of Shannon entropy (Xia et al. 2003) implemented in DAMBE5 (Xia 2013). The empirical proportion of invariant sites (P inv = 0.2) was estimated from the data. Xia's test for fully resolved sites, with the replicate number of 1,000 computed the index of substitution saturation (I ss = 0.205) for 32 operational taxonomic units, was substantially smaller than the critical index of substitution saturation (I ss.c = 0.684 assuming a symmetrical topology and I ss.c = 0.362 assuming an asymmetrical topology). In addition, the rates of transitions and transversions increased approximately linearly with the K2P genetic distance (Figure 4, Additional file 6), with transitions being higher than transversions (mean κ = 1.37, Wilcoxon one-tailed rank-sum test, W = 7,602,731, P < 2.2e − 16), which is expected for closely related species (Ong et al. 2011). Overall, saturation tests revealed no significant substitution saturation in our data, implying validity for phylogenetic reconstruction.
Both ML and Bayesian trees showed identical topologies for each clade ( Figure 5A,B). The monophyly of the group Ulidiinae + Otitinae was strongly supported (PP = 1.00, BS = 95), and most of the individual species branches within it were also well supported (PP > 0.95, BS > 75). The clade Physiphora Fallén, 1810 was sufficiently supported by ML (BS = 88) but not by Bayesian analysis (PP = 0.88). The monophyly of the group Timia Wiedemann 1824 + Ulidia Meigen, 1826 was not well supported by Bayesian analysis (PP = 0.90), and not by ML (BS = 34). The clade Otitinae was not recovered as monophyletic by Bayesian analysis (PP = 0.45) nor by ML (BS = 8).
The codon ML analysis resulted in an altered topology ( Figure 6A,B). Interestingly, however, the codon ML tree showed clustering patterns that reach an agreement with phylogeny corroborated by morphological observations. Overall, major clades on the codon ML tree generally received higher branch supports than on ML and Bayesian trees. We put a threshold value of 0.90 for aBayes support to be sufficient for recovering significant branches, as it is a more conservative metric than PP (Anisimova et al. 2011). The monophyly of the group Ulidiinae + Otitinae was strongly supported (aBayes = 1.00). The  paraphyletic relationship of the clade Physiphora to the group Ulidiinae (excluding Physiphora) + Otitinae was weakly supported (aBayes = 0.60). Within significant, the monophyletic (aBayes = 0.91) group Ulidia + Timia, Ulidia and some species of Timia formed a highly supported cluster (aBayes = 0.98). Although the branch Otitinae was also not significant on the codon ML tree, it received higher aBayes support (0.58) than original Bayesian PP (0.45). It is important to note that within Otitinae, Homalocephala species combined into a single cluster, but with low support (aBayes = 0.57). The genus Melieria received a very high aBayes support (0.98) implying its monophyly.
Deeper branches mostly recovered with high supports (aBayes > 0.9) as on ML and Bayesian trees.

Morphological analysis
We examined external morphology of approximately 6,800 fly specimens belonging to 82 species of 3 genera in the subfamily Ulidiinae and 2,700 fly specimens from 107 species of 12 genera in the subfamily Otitinae. Based on this analysis, we confirmed morphological characters that are able to separate all studied genera. In particular, the genital structures (surstyles, phallus, and phallus apodeme) were shown to be the most useful for distinguishing species. We moved the results of the morphological analysis to the 'Discussion' section for easier perception.

Discussion
Our phylogenetic analysis suggests the existence of three major clusters -Physiphora (Ulidiinae), Timia + Ulidia (Ulidiinae), and Otitinae + Homalocephala (Figures 4 and  5B). Inclusion of Homalocephala into Otitinae coincides with Hennig's classification [1], whom previously considered it within Ulidiidae. The subfamily Ulidiinae (Physiphora + (Timia + Ulidia)) forms a basal paraphyletic group (BS = 33, PP = 0.64, aBayes = 0.60) nested between P. fimetaria and Otitinae. Nevertheless, morphologically these genera are similar in having a bare vein R 1 , as well as in genital structures (non-rupted closed hypandrium, well-developed parameres, a bare distiphallus (rarely with a group of setae in the center of distifallus) with membranous lobes, and a phallus apodeme with lateral lobes). Thus, there might be symplesiomorphic similarity. The subfamily Otitinae occupies the apical position on all of the phylogenetic trees (Figures 5 and 6). Despite this, the lineage was not statistically recovered as a monophyletic group, with low support values (BS = 8, PP = 0.45, aBayes = 0.58). Morphologically, the monophyly of Otitinae (without Homalocephala) was supported by a spinulose phallus, absence of the glans, absence of membranous lobes on the distiphallus and by the presence of the female sternal apodemes. We assume these characters to be an apomorphic similarity. Interestingly, the Homalocephala genus takes intermediate position between Ulidiinae Otitinae, which can be corroborated by previous studies (including Hennig 1973a, b). Recovery of basal and subbasal positions of the Seioptera species within Otitinae is partially supporting the hypothesis of Kameneva and Korneyev (1994), which suggests inclusion of the Homalocephala and Seioptera genera into the Seiopterini tribe, however, within Ulidiinae. Our results confirm the uniting of Ulidiinae and Otitinae into a single family.
Physiphora (mostly the Afrotropical genus, 24 species) branches out as a monophyletic clade with relatively high support (BS = 88, PP = 0.88, aBayes = 1.00). This observation is in agreement with morphological data about the genus. All Physiphora species possess a very similar structure of exoskeleton and genital characters: Physiphora always have transparent wings ( Figure 7X,Y); spermathecae covered by small papillae, the shape of the phallus tip (glans) in combination with a bare distiphallus with membranous lobes; and surstyles without prensisetae ( Figure 8W,X). The genus Physiphora is distributed mostly in the Old World, but includes several cosmopolitan species with the worldwide area. Thus, early divergence might have allowed Physiphora species to spread across such a vast range.
Ulidia (Palearctic genus, 21 species) and Timia (Palearctic genus, 59 species) most likely form a monophyletic (BS = 34, PP = 0.90, aBayes = 0.91) cluster, although these two genera do not create particular clusters within it. The morphological data that closely relates these two genera, the overall external characters and genital structures and having little or no hiatus, firmly supports the phylogenetic inference. Group specialists often encounter issues in separation between Ulidia and Timia, so the possible merging of these genera has been suggested. The differences used for separating Ulidia and Timia so far are mainly the following: the frons is smooth (in Timia) versus dimpled (in Ulidia) (with some exceptions); the head and thorax are microtrichose (in Timia) versus bare (in Ulidia, but some assigned to Timia have a shiny head and thorax, and the frons of Ulidia metope Kameneva, 2010 is widely microtrichose) (Chen and Kameneva 2009;Kameneva 2010). The monophyletic status of the group Timia + Ulidia is also confirmed by the distribution of the species with a shared ecological niche -the majority of Ulidia and Timia occur in arid and semiarid areas of the Palaearctic. Based on these data, it is more likely that the formal characters defined Ulidia and Timia, unambiguously identifying just a fraction of species in each genus. Likewise, the subgenera Timia and Empyelocera in the genus Timia seem to be unnatural taxa, which is only differentiated by width of parafacial area (Zaitzev 1982). This assumption is congruent with our hypothesis that the genera Ulidia and Timia have recently diverged and are currently on a course of differentiation. Within these groups, typical representatives as well as transitional forms can be found. The molecular phylogenetic reconstruction does not exclude a more complicated scenario such as Ulidia and Timia not being natural groups but forming a cluster of several subgroups connected with multiple transitions. Timia have transparent wings, some species have dark basal costal (bc), costal (c), and subcostal (sc) cells and with apical wing spot. A basal bcu cell has a well-developed posteroapical projection, whose length has an identification value ( Figure 7Z, A1,A2,A3). Male surstyles do not possess strong prensisetae, occasionally with setae on the inner lobe ( Figure 8Y,Z, A1,A2). The genus Ulidia possesses a poorly developed wing pattern, the species Ulidia megacephala, Ulidia melampodia Loew, 1873, Ulidia salonikiensis Hennig, 1940, Ulidia kandybinae Zaitzev, 1982, Ulidia splendida Zaitzev, 1982, Ulidia nigricubitalis Zaitzev, 1982, and Ulidia transcaspica Galinskaya, 2011 exhibit the greatest development of the character: dark costal, subcostal, and (sometimes) bcu cells and an apical wing spot ( Figure 7A4,A5). Ulidia possess three smooth spermathecae with two of them having a collective sperm duct. Male genitalia has the shape of the phallus tip (glans) in combination with a bare distiphallus (rarely with a group of setae in the center of the distiphallus) and distiphallus with membranous lobes. Male surstyles do not possess strong prensisetae, occasionally with setae on the inner lobe ( Figure 8A3,A4). It is important to note that within Timia, we found three species (Timia albifacies Gorodkov and Zaitzev, 1986, Timia gussakovskyi Gorodkov and Zaitzev, 1986, and Timia komarowi Mik, 1889) that have the distiphallus in the middle part with an area densely covered by relatively strong setae. We hypothesize that setae on the distiphallus of Timia and Otitinae emerged in parallel, possibly having unique origins.
Homalocephala species (nine species total: four Holarctic, four Palaearctic, and one Nearctic) have always been included into the clade Otitinae with different, lowly supported clustering patterns on the Bayesian/ML tree ( Figure 5A) and codon ML tree ( Figure 6A). On the codon ML tree, the Homalocephala cluster likely exhibits sister relationships (aBayes = 0.84) with the Otites + Tetanops + Herina + Ceroxys + Melieria group (aBayes = 0.85). Homalocephala species possess the following relevant morphological characters: no glans on the distiphallus, two katepisternal setae, and four spherical spermathecae. This relates these species to the Seioptera and Pseudoseioptera genera, whereas having vein R 1 covered with setulae along its full length relates Homalocephala to Pseudoseioptera. We notice that Homalocephala and Seioptera genera were placed basally or subbasally relatively to Otitinae ( Figures 5A and 6A). Species of Homalocephala possess transparent wings, occasionally with apical spots (on apexes of veins R 2 + 3 and R 4 + 5 ) or with medial spots occupying pterostigma and sometimes extending downward. The bcu cell is enclosed with arcuate vein without breaking ( Figure 7T,U,V,W). Male surstyles possess figured prensisetae ( Figure 8S,T,U,V) that are typical only of genera Homalocephala, Pseudoseioptera, and Ceroxys and are probably the result of parallel origin or convergent evolution. Interestingly, the areal of Homalocephala partially coincide with the areal of Pseudoseioptera inhabiting Holarctic temperate forests; moreover, both of their larvae live under the bark of falling trees. The position of this genus is disputed, with differerent scientists including it in Otitidae, Ulidiidae, or Pterocallidae (Kameneva and Korneyev 1994;Hendel 1910aHendel , b, 1928Stackelberg 1933;Hennig 1939Hennig , 1940Richter 1970a, b;Soós 1984;Zaitzev 1984;Krivosheina and Zaitzev 1989;Greve 1998;Nartshuk 2003;Curran 1934a, b;Steyskal 1965Steyskal , 1977McAlpine 1977McAlpine , 1989Rohdendorf 1977;Evenhuis 1989;Frey 1964) (Figure 1). Based on phylogenetic inferences about the genus, Homalocephala (Figures 5A and 6A) tend to be placed within the Otitinae clade, though the taxonomic status of the genus needs to be extensively revised. Morphologically, the Homolocephala lineage remains unresolved too, as most of the characters examined are plesiomorphic and cannot reliably relate it to either Otitinae or Ulidiinae.
Melieria contains 38 species: 2 Holarctic, 4 Nearctic, and 29 Palaearctic. In our study, Melieria omissa (Meigen, 1826), Melieria cana, Melieria acuticornis (Loew, 1854), and Melieria crassispennis (Fabricius, 1794) exhibited a tendency to form a single cluster with an apical topology ( Figures 5A and 6A). While the codon ML tree gave the genus high branch support (aBayes = 0.98), the Bayesian/ ML tree ambiguously clustered it with Ceroxys munda (BS = 10, PP = 0.51). Interestingly, Melieria shows maximal levels of intraspecific diversity, whereas interspecific comparisons within the genus attain the minimal distances. Perhaps, these results suggest deep conspecific diversity of M. cana, which in turn may indicate a current process of speciation. Species of Melieria possess an R 1 vein with setulae in its apical portion and a bcu cell with very short posteroapical extension ( Figure 7H,I,J,K). The male distiphallus is covered by long bristles, lacks membranous lobes, and lacks a glans; it also has surstyles with two to five prensisetae ( Figure 8G,H,I,J).
The single representative of Tetanops Fallén, 1820 (Holarctic genus with eight Nearctic and six Palaearctic species) -Tetanops sintenisi Becker, 1909 -was placed with Otites formosa on the Bayesian/ML tree with weak support (BS = 15, PP = 0.69), but on the codon ML tree, it was placed within the lowly supported clade containing Herina lugubris, Ceroxys hortulana, and Otites lamed (aBayes = 0.77). T. sintenisi possesses an R 1 vein with setulae and a bcu cell with short posteroapical extension ( Figure 7S). The male distiphallus is covered by long strong bristles and lacks membranous lobes and a glans; it has surstyles with two to eight prensisetae on the medial lobe ( Figure 8R).
Herina R.-D., 1830 has 31 species: 19 Palaearctic species, 4 Nearctic, 6 Oriental, and 2 Neotropic. Herina lugubris (Meigen, 1826) and Herina frondescentiae (L., 1758) were differently clustered with low branch support values ( Figures 4A and 5A). Species of Herina possess an R 1 vein with setulae and a bcu cell with short, often indistinct posteroapical extension ( Figure 7F,G). The male distiphallus is covered by short strong bristles and lacks membranous lobes and a glans; it has surstyles with prensisetae and its lobes are often elongated and modified ( Figure 8E,F). Herina are generally widespread in the majority of zoogeographic regions, excluding the Afrotropic (Kameneva and Korneyev 2006).
Representatives of Otites Latr., 1804 (31 species: 5 Nearctic and 26 Palaearctic species) -Otites guttata (Meigen, 1830), Otites lamed (Schrank, 1781), and Otites formosa were scattered across all the trees showing no particular clustering patterns ( Figures 5A and 6A). Species of Otites possess an R 1 vein with setulae and a bcu cell with short posteroapical extension ( Figure 7N,O,P). The male distiphallus is covered by strong bristles and lacks membranous lobes and a glans. The medial lobes of the surstyles are covered by an abundance of prensisetae (15 to 40), which clearly represent the genera's character ( Figure 8M,N,O). Otites is the only Holarctic genus, with the majority of the other species distributed in Southern Europe.
The single representative of Seioptera (two species: one Holarctic and one Nearctic) -S. vibrans (Linnaeus, 1758) -was nested within the lineage Myennis on the codon ML tree (aBayes = 0.72; Figure 6A). However, on the Bayesian/ML tree, it showed different topology clustering with Myennis sibirica (BS = 28, PP = 0.57; Figure 5A). Morphologically, this genus is close to Pseudoseioptera and Homalocephala based on the following characters: no glans on the distiphallus, two katepisternal setae, and four spherical spermathecae. Species of Seioptera possess a bare R 1 vein, a cup cell with short posteroapical extension, and an sc cell darkened and an apical spot ( Figure 7R). The male distiphallus lacks a glans and membranous lobes and is covered by relatively strong bristles along its full length; it has surstyles with two big prensisetae on the medial lobe ( Figure 8Q).
Representatives of Myennis R.-D., 1830 (four Palaearctic species) -Myennis octopunctata (Coquebert, 1798) and Myennis sibirica Portschinsky, 1892 -formed the single cluster with non-significant branch support on the codon ML tree (aBayes = 0.72; Figure 6A). No particular clustering was observed on the Bayesian/ML tree. Species of Myennis possess an R 1 vein with setulae and a bcu cell with short posteroapical extension ( Figure 7L,M). The male distiphallus is covered by short strong bristles and lacks membranous lobes and a glans; the lateral lobes of the surstyles have two to three prensisetae ( Figure 8K,L). The genus Myennis is mostly restricted to forest zones. Myennis and Seioptera genera form a singe branch on all the phylogenetic hypotheses.
The single representative of Pseudotephritis Johnson, 1802 (five species: two Palaearctic and three Nearctic) -Pseudotephritis corticalis (Loew, 1873) -is not consistently placed on all of the trees. Species of Pseudotephritis possess an R 1 vein with setulae and a bcu cell with long posteroapical extension ( Figure 7Q). The male distiphallus is covered by strong bristles and lacks membranous lobes and a glans; it has surstyles with prensisetae ( Figure 8P).
In summary, based on phylogenetic inference, the subfamily Otitinae did not form a statistically supported monophyletic group. However, we found morphological synapomorphic characters that may unambiguously distinguish Otitinae from Ulidiinae, namely a spinulose phallus, the absence of the glans, the absence of membranous lobes on the distiphallus, and the presence of the female sternal apodemes. Presumably, the COI region has no sufficient phylogenetic signal to resolve genera divergence.

Conclusions
The phylogenetic analysis of COI showed satisfactory results in recovering some major lineages of the Ulidiidae family. On all phylogenetic trees obtained in our survey, there is a clear tendency of clustering at higher levels with separation of (Physiphora + (Timia + Ulidia)) (Ulidiinae) and (Otitinae + Homalocephala). Otitinae take apical position within the Ulidiidae family. However, monophyly of the Ulidiinae subfamily was not supported, forming basal paraphyletic group. Our tree inferences provide evidence that the Homalocephala genus has ambiguous evolutionary relationships with other Ulidiinae and Otitinae that was also shown by morphological analyses. In addition, our results support the inclusion of the Homalocephala and Seioptera genera into the Seiopterini tribe that was initially proposed by Kameneva and Korneyev (1994).
Additional material from the subfamilies Ulidiinae and Otitinae would aid to further investigation to confirm the phylogenetic status of Ulidiinae, Otitinae, and the genus Homalocephala.
We also conclude that analysis of other molecular markers may help to support the hypothesis of an evolutionary split having occurred between Ulidiinae and Otitinae. Nevertheless, phylogenetic reconstruction based on Markov models of codon evolution provides a good resolution for our limited dataset. Most likely, the codon models coupled with the ML approach are capable of capturing more general phylogenetic patterns, making them very promising and attractive tools for modern systematic entomology.