Molecular phylogenetics reveals novel relationships within Empidoidea (Diptera)

Empidoidea represent a large and diverse superfamily of true flies, and to date no stable hypothesis on the phylogeny exists. Previous classifications have been based on morphological data and the relationships among several groups are still unknown. Using the mitochondrial genes cytochrome oxidase c subunit I (COI) and cytochrome β (Cytβ) and the nuclear genes carbomoylphosphate synthase domain of rudimentary (CAD), elongation factor‐1α (EF‐1α) and isocitrate dehydrogenase (IDH) in a Bayesian analysis, we tested the support of higher taxonomic groups within this large superfamily of flies. We re‐evaluated previous hypotheses of evolution within the group and present a highly supported phylogenetic hypothesis. Atelestidae, Dolichopodidae, Empididae and Hybotidae were supported as monophyletic families, with Atelestidae as sister group to the remaining Empidoidea. Within the family Hybotidae, Bicellariinae stat.n. formed the sister group to the other subfamilies. The family Ragadidae stat.n. is established to include the subfamily Ragadinae and the new subfamily Iteaphilinae subfam.n.; Ragadidae was sister group to the Empididae. Dolichopodidae was found to form a sister group to Ragadidae plus Empididae. Within Empididae, Hemerodromiinae was found to be a nonmonophyletic group. The tribes Hilarini and Hemerodromiini stat. rev. were recovered as sister groups, as were Empidini and Chelipodini stat. rev. The former family Brachystomatidae was found to be nested within Empididae. A revised classification and diagnoses of nondolichopodid families, subfamilies and tribes are provided.


Introduction
The superfamily Empidoidea Latreille (Insecta: Diptera) includes over 10 000 described species (Pape et al., 2011) and is represented in all biogeographical regions except Antarctica. According to fossil records (Grimaldi & Engel, 2005) the group existed in the Jurassic, with families and subfamilies already diversified in the Early Cretaceous (Grimaldi & Cumming, 1999). Several species have gained interest from biologists owing to elaborate mating rituals (e.g. representatives of the genus Empis Linnaeus), and from ecologists due to freshwater life histories (e.g. some Clinocerinae). A vast diversity in morphology in combination with being closely related to the extremely diverse Cyclorrhapha has sparked a number of hypotheses about the evolution of the group. The internal classification of Empidoidea has been subject to a number of revisions and discussion since Brauer (1883) first considered Correspondence: Emma Wahlberg, Department of Zoology, Swedish Museum of Natural History, P. O. Box 50007, SE-104 05 Stockholm, Sweden. E-mail: emma.wahlberg@nrm.se that Empididae Latreille and Dolichopodidae Latreille were two closely related families. Chvála (1983), Wiegmann et al. (1993), Collins & Wiegmann (2002) and Sinclair & Cumming (2006) all provided detailed historical reviews of the superfamily. The two relatively recent classifications that have gained most acceptance are those by Chvála (1983) (Fig. 1A) and Sinclair & Cumming (2006) (Fig. 1B). Even though both have been used simultaneously, the latter has taken over as the dominant classification scheme. In the analysis by Sinclair & Cumming (2006), only parts of the phylogeny of Empidoidea were resolved. Two more recent publications by Yang et al. (2006) and Yang et al. (2007) use an older classification, dividing Empidoidea in two families (Dolichopodidae and Empididae), with several rearrangements. All previous classifications and most phylogenetic hypotheses comprising Empidoidea have been based on morphological data. Tentative studies producing a phylogenetic hypothesis based on molecular data were first carried out by Collins & Wiegmann (2002) using 28S ribosomal DNA and elongation factor-1 (EF-1 ). Their phylogeny corroborated Chvála's (1983) hypothesis, but with low support. They were unable to resolve within-family relationships. Later Moulton & Wiegmann (2004) explored a large part of a region of the CAD gene in phylogenetic reconstruction of the superfamilies of the Eremoneura. They produced high support for many higher groups, but taxon sampling was insufficient to draw conclusions regarding relationships below family level. The same authors continued their exploration by adding 28S ribosomal DNA to the analysis (Moulton & Wiegmann, 2007) (Fig. 1C). Support for major groups, in favour of the family classification by Chvála (1983), was increased compared with their previous study. However, questions regarding relationships within families and among tribes remained and increased sampling of both taxa and molecular data was suggested. In this paper we test support for the monophyly of families, subfamilies and tribes, with a focus on the families Empididae and Hybotidae Meigen.

Material and methods
All currently recognized empidid and hybotid families and tribes, as well as several incertae sedis taxa sensu Sinclair & Cumming (2006) were included in the analysis, represented by 83 ingroup taxa (Table S1). An additional six taxa from both lower Brachycera and Cyclorrhapha were selected to serve as outgroups. The samples were made available mainly from the Swedish Malaise Trap Project (SMTP) (Karlsson et al., 2005), and staff and associates of the Swedish Museum of Natural History (NHRS). All material has been stored at approximately −20 ∘ C. As ground-living and aquatic species were rare in the Malaise trap samples, additional collection methods such as using sweep nets, pit fall traps and light traps were carried out to increase the taxonomic coverage. Some specimens in the subfamily Clinocerinae were donated to the project by Dr Marija Ivković (University of Zagreb, Croatia). Gene fragments from each of mitochondrial cytochrome (Cyt ) and cytochrome oxidase c subunit I (COI), and the nuclear carbomoylphosphate synthase domain of rudimentary (CAD), elongation factor-1 (EF-1 ) and isocitrate dehydrogenase (IDH) were included in the analysis. The primers used are listed in Table S2. Difficulties in amplifying Cyt in several taxa were solved by exchanging a few nucleotides to inosine in the primer sequence (Table S2). A new set of primers for amplifying EF-1 were designed (Table S2) for the same reason, using previous published sequences in GenBank (Clark et al., 2015) of Empidoidea and close relatives. Photography of morphological characters were carried out on specimens after lysis, and produced with manual focus stacking, using a Nikon DS-Fi1 camera on a Nikon Eclipse 80i microscope. Photos were automatically aligned and stacked in helicon focus 6 (Helicon Soft Ltd., Dominica) and were thereafter edited and finalized in Adobe photoshop cc 2017.1.1.

Extraction and sequencing
Following the manufacturer's protocols, the KingFisher™ Duo (Thermo Scientific, Waltham, MA, U.S.A.) extraction robot was used together with KingFisher™ Cell and Tissue DNA Kit (Thermo Scientific) to extract DNA. For large specimens the abdomen was removed from the body and used for extraction, and thereafter kept together with the rest of the body. For small specimens, the whole body was used for extraction. Lysis was performed in 56 ∘ C overnight. All extracted material is kept in 80% ethanol as vouchers at the Swedish Museum of Natural History (SMNH). The PCR reactions were performed using Ready-To-Go PCR Beads (Amersham Biosciences, U.K.). The mixture for PCR reactions was set up with 21 L H 2 O, 1 L each of reverse and forward primer and 2 L DNA extract, except for when amplifying IDH and EF-1 , where 3 L of DNA extract was required to obtain best results. PCR reaction protocols were optimized regarding temperature and number of cycles for each primer pair and are listed in Table S3. PCR products were cleaned using Exo-Fast (QiaQuick PCR Purification Kit; Qiagen, Inc., Valencia, CA, U.S.A.) and sequencing reactions were carried out by Macrogen Inc. (South Korea) using the same primer pairs as for PCR reaction. The resulting gene fragments were assembled into contigs in geneious 8.1.8 (Kearse et al., 2012) where they were inspected and primer sequences removed. The raw sequences were thereafter stored in voseq 1.7.4 (Peña & Malm, 2012), which was also used for managing sequences and creating datasets for analyses. To detect contamination, sequences were compared with known sequences of relatives as well as the GenBank database using blast (Madden, 2013).

Alignment and partitioning
Sequences were aligned using mafft 7.017 (Katoh et al., 2002) implemented in geneious 8.1.8. The mitochondrial genes COI and Cyt were aligned with the G-INS-1 method, while the nuclear genes CAD, Ef1-and IDH contained regions of gaps and were aligned using the E-INS-i method. In both cases, sequence direction adjustment was selected while other options were left at default. Long gaps or highly divergent regions generate potential alignment problems in the nuclear genes and were removed using the software gblocks 0.91b (Castresana, 2000;Talavera & Castresana, 2007) under the default (stringent) options setting. Substitution saturation for codon 1+2 and 3 was assessed using Xia's method (Xia et al., 2003;Xia & Lemey, 2009) in dambe 6.4.29 (Xia, 2013), an entropy-based index of substitution saturation.

Phylogenetic analysis
Bayesian analysis was carried out using mrbayes 3.2.6 (Huelsenbeck & Ronquist, 2001;Ronquist & Huelsenbeck, 2003) on the CIPRES Science Gateway (Miller et al., 2010). Following the results from the saturation assessment, datasets for analysis were set up using a partitioned scheme with codon positions 1+2 and 3 for each gene assigned to separate subsets. Rate variation across sites was set to gamma distribution with a proportion of invariable sites. To integrate over substitution model space, a mixed setting was used (nst = 'mixed')  ( Huelsenbeck et al., 2004). Temperature was set to 0.11 and number of chains to 4, and two parallel runs were performed. The analysis ran for 100 million generations and a sampling frequency of 10 000. The average standard deviation of split frequencies was used as a first measure of convergence (< 0.005), and secondly using the log files in tracer 1.6  for convergence and effective sample size (ESS). The ESS was, for all parameters, at least 500 for separate runs and at least 1300 combined. The majority rule tree was generated after removing the first 25% burn-in trees. The majority rule consensus tree was viewed in and exported from figtree 1.4.2 (Rambaut, 2014), and annotated and finalized in Adobe illustrator cc 19.2.1.
The saturation test using the ratio of observed entropy to the entropy of full substation saturation is summarized in detail in Table S4. First and second codon positions show little saturation in all genes (index of substitutional saturation (Iss) < critical index of substitutional saturation (Iss.c), P < 0.05). The third codon position of the nuclear genes CAD, EF-1 and IDH show little saturation assuming a symmetrical topology (Iss < Iss.c, P < 0.05), but are very poor for phylogenetics assuming an asymmetrical topology (Iss > Iss.c, P > 0.05). The third codon position of COI shows saturation for both, assuming a symmetrical (Iss < Iss.c, P > 0.05) and asymmetrical (Iss > Iss.c, P < 0.05) topology, while the same position for Cyt shows saturation assuming an asymmetrical topology (Iss > Iss.c, P < 0.05). Based on these results, showing saturation of the third codon position in all genes, the third codon position was partitioned and modelled independently in the Bayesian analysis.
The majority rule consensus tree from the Bayesian inference showed strong support for all higher taxa (Fig. 2). Empidoidea formed a monophyletic superfamily with Atelestidae as the sister group to all other empidoid groups. The family Hybotidae was sister group to the remaining Empidoidea, excluding Atelestidae. Within Hybotidae, the subfamily Hybotinae, herein represented by the genera Bicellaria Macquart and Hybos Meigen, appeared as polyphyletic. Bicellaria was recovered as sister group to all other Hybotidae, while Hybos was the sister clade to Oedaleinae. Tachydromiinae was recovered as a monophyletic subfamily, with Drapetini as sister group to Symballophthalmini and Tachydromiini. Empididae was recovered as paraphyletic, with Brachystomatidae Melander as sister to the empidid subfamilies Empidinae and Hemerodromiinae. Ragas Walker and Iteaphila Zetterstedt formed a monophyletic group sister to Empididae and Brachystomatidae. Clinocerinae was recovered as monophyletic and both Empidinae and Hemerodromiinae were found to be nonmonophyletic: Hemerodromiini and Hilarini formed the sister group to Empidini and Chelipodini. At the generic level there were also incongruences compared with current classification schemes: Anthepiscopus Becker was nested within Iteaphila, and in Empidini the genus
Rhamphomyia Meigen was nested within Empis. The sister group to the Empididae and the Ragas and Iteaphila groups was Dolichopodidae, with Microphorinae as the sister group to the remaining dolichopodids.

Phylogenetic relationships and taxonomic implications
The results from the analyses confirm that many groups in the classification by Sinclair & Cumming (2006) (Fig. 1B) are monophyletic. Hennig (1970) first proposed that Atelestinae is closely related to Hybotidae and Microphoridae Collin. Chvála (1983), however, raised the group to family level and placed it as sister group to the Cyclorrhapha. Wiegmann et al. (1993) detected a close relationship between Atelestidae and Hybotidae in their analysis using morphological characters, and Sinclair & Cumming (2006) found Atelestidae as the sister group to Hybotidae. However, the finding that Atelestidae form the sister group to all remaining Empidoidea is coherent in all studies using molecular data (Collins & Wiegmann, 2002;Moulton & Wiegmann, 2004) (e.g. Fig. 1C), including the present analysis.
The position of Bicellaria near the root of Hybotidae supports Moulton & Wiegmann (2007) (Fig. 1C) when using combined CAD and 28S gene data. Chvála (1983) (Fig. 1A) placed Bicellaria within Trichini in Ocydromiinae, while Sinclair & Cumming (2006) created the tribe Bicellariini and included it in Hybotinae. Based on the results herein we suggest that Bicellariini form a group outside Hybotinae and raise it to subfamily Bicellariinae stat.n.
Brachystomatidae comprise the three subfamilies Ceratomerinae, Trichopezinae and Brachystomatinae as combined by Sinclair & Cumming (2006), who also raised it to its present status as family. Cumming et al. (1995) hypothesized a close relationship of this group to Dolichopodidae + Microphorinae. This was corroborated in the results by Sinclair (1995). Molecular analysis by Collins & Wiegmann (2002), however, did not support that Brachystomatidae represent the sister group to Dolichopodidae + Microphorinae, results further corroborated by Moulton & Wiegmann (2007). Even though the sample size is too small and restricted to Trichopezinae to investigate the monophyly of Brachystomatidae, maintaining family-level status for Brachystomatidae would necessitate extensive changes in the classification of the remaining empidids with several clades raised to family rank as well. We feel that this would create an overly split classification, so we choose a conservative approach and move this group into Empididae.
Empidinae and Hemerodromiinae have previously been recognized as sister taxa, due to the presence of heavy sclerotized male cerci with clasping abilities in species of both groups (Ulrich, 1975;Chvála, 1983;Cumming et al., 1995). Hemerodromiinae is distinct from Empidinae by the presence of a long, tubular fore coxae and raptorial forelegs widely separated from the middle and rear pairs. Hemerodromiini and Chelipodini were formalized by Melander (1947). Wiegmann et al. (1993) excluded Chelipodini and found that Hemerodromiini were most closely related to Brachystomatinae + Clinocerinae. Cumming et al. (1995), on the other hand, found that Hemerodromiinae is most closely related to Empidinae because they share the same male cercal morphology, but the authors did not include characters to the test the monophyly of the subfamilies. Collins & Wiegmann (2002) included one taxon from each tribe and recovered Hemerodromiinae as paraphyletic yet closely related to Empidinae. The close relationship to Empidinae was also confirmed by Sinclair & Cumming (2006), on the basis of the presence of an upcurved phallus and widely separated epandrial lamellae in the two groups. They also recovered a monophyletic Hemerodromiinae; however, their representative of Chelipodini, Chelipodozus Collin, is problematic in the sense that it differs from the other genera in the tribe by having a bare laterotergite (Plant, 2008). This state is considered a synapomorphy for Hemerodromiini.
Based on morphological characters, Plant (2011b) found Chelipodozus nested within Hemerodromiini and Chelipodini genera Afrodromia Smith and Drymodromia Becker as near-basal taxa, preceding the remaining Chelipodini + Hemerodromiini. Moulton & Wiegmann (2007) recovered Hemerodromiinae as a monophyletic clade based on an analysis using fragments of CAD and 28S. While Hemerodromiinae formed the sister group to the Empidinae, it rendered Hemerodromiini paraphyletic. Based on our study including multiple taxa from both groups, we include the Hemerodromiini and Chelipodini as tribes in the Empidinae.
The Ragas group, as proposed by Sinclair (1999a) and supported by Sinclair & Cumming (2006) (including the genera Dipsomyia Bezzi, Hormopeza Zetterstedt, Hydropeza Sinclair, Ragas and Zanclotus Wilder), and the Iteaphila group (including Anthepiscopus and Iteaphila) have previously been grouped together with Oreogeton Schiner, Gloma Meigen and Hesperempis Melander in the subfamily Oreogetoninae (Chvála, 1976). Chvála (1983) hypothesized this subfamily to be a primitive group within Empididae. The results here do not support Oreogetoninae sensu Chvála (1976), but instead corroborate the results by Wiegmann et al. (1993), Collins & Wiegmann (2002), Sinclair & Cumming (2006) and Moulton & Wiegmann (2007). Sinclair (2016) erected the subfamily Ragadinae for the Ragas group but did not discuss its phylogenetic placement other than keeping it incertae sedis within Empididae sensu Sinclair & Cumming (2006). The results here show that the Ragas and Iteaphila clade is sister group to remaining empidids, similar to the results by Moulton & Wiegmann (2007), where Hormopeza was also recovered within the clade. Based on diagnostic characters as well as the genetic differences separating this group we recognize the Ragadinae and Iteaphila group as distinct subfamilies within the family Ragadidae stat.n.

New classification and diagnoses in Empidoidea
Based on our new phylogenetic hypothesis, a revised classification of the Empidoidea is presented below, with subfamilies and tribes shown for all families except Dolichopodidae. The genera Gondwanamyia Sinclair, Homalocnemis Philippi, and Oreogeton have been placed incertae sedis within the superfamily and remain unplaced until further studies resolving the placement of these are clarified. A simplified cladogram based on our analysis showing the proposed classification and hypothesized relationships is presented in Fig. 3. Morphological characters applied in the diagnoses for the various taxa are mainly derived from Chvála (1983) and Sinclair & Cumming (2006). The terminology follows Cumming et al. (1995) for male terminalia, McAlpine (1981) for adult morphology, and Stuckenberg (1999) for antennae. Atelestinae Hennig, 1970: 1. Type genus Atelestus Walker.
Diagnosis. Similar to Empididae and Ragadidae by the presence of symmetrical male terminalia without rotation (Fig. 5A) and origin of vein R s at a distance from humeral crossvein (h) as long or longer than length of h (Fig. 6A). Distinguished from Ragadidae by having a costa ending at or near M 1+2 , and from Empididae by having prosternum separated from proepisternum (Fig. 7A).
Comments. Although this group has consistently been closely associated with Hybotidae in morphological studies, the molecular evidence showed that it formed the sister group to all remaining groups in the superfamily. Atelestidae was formally expanded to include the genus Nemedina Chandler as the single genus in the subfamily Nemedininae Sinclair & Cumming.

Atelestinae Hennig
Diagnosis. Separated from Nemedininae by the presence of a straight r-m and R s not distinctly short (Fig. 6A).
Comments. This group have been considered a monophyletic group based on morphology by both Chvála (1983) and Sinclair & Cumming (2006).

Diagnosis. Distinguished from Atelestinae by having arched vein r-m and a short R s .
Comments. Including only one extant genus, but several fossil taxa have been described as closely related (Grimaldi & Cumming, 1999;Sinclair & Shamshev, 2003). (Sinclair & Cumming, 2006). Diagnosis. Male genitalia rotated, distinguished from Hybotidae by being dextrally rotated between 90 ∘ and 180 ∘ , including segment 8 and sometimes also segment 7 (Fig. 5B). Wings with simple R 4+5 (Fig. 6B); costa ending near or at M 1 /M 1+2 or continuing along the wing margin; and the point of origin of R s is at, or very close to, h. Prosternum often separated from proepisternum (Fig. 7B).

Dolichopodidae Latreille
Comments. Chvála (1983) established the family Microphoridae, but the validity of this was subsequently questioned and the status was changed to subfamily within the Dolichopodidae (Sinclair & Cumming, 2006;Germann et al., 2011). In our analysis the monophyly of Dolichopodidae as well as the classification of Microphorinae as subfamily within the Dolichopodidae were corroborated. The phylogeny within the Dolichopodidae was outlined based on morphological and molecular data in several studies, including Brooks (2005)  Empididae Giebel (1856: 206).
Diagnosis. Distinguished from the Dolichopodidae and Hybotidae by the presence of unrotated and symmetrical terminalia (Fig. 5C). The point of origin of R s at a distance from the humeral crossvein h equal to or longer than length of h (Fig. 6C). Distinguished from the Atelestidae by the presence of a costa that ends at, or just beyond, R 4+5 , or continues along the wing margin. Except for Niphogenia Melander spp. and Philetus Melander spp., distinguished from both Ragadidae and Atelestidae by the presence of a prosternum that is fused with proepisternum and forming a precoxal bridge (Fig. 7C).
Comments. Brachystomatidae is here brought back as a subfamily within Empididae and the subfamilies of the former family are consequently lowered to tribes. The incertae sedis genera Brochella Melander, Dryodromia Rondani, Hesperempis Melander and Philetus were not included in our analysis and remain unplaced until further studies confirm their placement.
Diagnosis. Variable group with wing with or without anal lobe; CuA 2 joining the A 1 before crossvein bm-cu (Fig. 6H) in Ceratomerini and Trichipezini, beyond in Brachystomatini. Scape at most slightly longer than the pedicel, twice as long in Ceratomerini; conus of the pedicel absent (Fig. 7K) in Brachystomatini and Trichopezini, present in Ceratomerini. Male cercus developed but simple in structure (cf. Fig. 5F). Females with acanthophorites on tergite 10 (Fig. 5J).
Comments. The group was originally treated as a subfamily of Empididae, but was elevated to family level by Sinclair & Cumming (2006). The status as a subfamily is re-established here. The tribes within this subfamily have previously been included as subfamilies within Empididae, but were excluded from the Empididae by Sinclair & Cumming (2006).

Brachystomatini Melander stat.n.
Diagnosis. Narrow wings as Ceratomerini, but distinct in in the vein CuA 2 that is joining A 1 beyond the crossvein bm-cu. Scape at most slightly longer than the pedicel, never twice as long as in Ceratomerini, and pedicel without conus (cf. Figs. 7I, J). Male cercus is articulated but simple in structure. Female tergite 10 is divided into a pair of lobes with stout spine-like setae (acanthophorites) (cf. Fig. 5J).
Comments. According to Sinclair (1995) the tribe comprises the two genera Anomalempis Melander and Brachystoma Meigen. When Sinclair & Cumming (2006) raised Brachystomatinae to family level, the nominate subfamily was defined to include Xanthodromia Saigusa as a third genus.
Diagnosis. Narrow wings and absence of an anal lobe (cf. Fig. 6D); and CuA 2 joins A 1 before the crossvein bm-cu. Scape twice as long as, or longer than, the pedicel; pedicel with a fingerlike conus. Male cercus well developed but simple in structure (cf. Fig. 5F). Female tergite 10 with acanthophorites (cf. Fig. 5J).
Diagnosis. Wings with well-developed anal lobe; and CuA 2 joining the A 1 before crossvein bm-cu (Fig. 6H). Scape at most slightly longer than the pedicel, twice as long in a few species, the conus of the pedicel absent (Fig. 7K). Male cersus simple (cf. Fig. 5F). Females with acanthophorites on tergite 10 (Fig. 5J).
Comments. According to previous analyses based on morphology, as well as our molecular analysis, this subfamily is monophyletic and forms the sister group to all other subfamilies in Empididae.  (Sinclair, 1995(Sinclair, , 1999b(Sinclair, , 2008. Diagnosis. Characteristic in having a well-developed anal lobe; the CuA 2 is joining A 1 before crossvein bm-cu (Fig. 6C); scape usually only slightly longer than, rarely twice as long as, pedicel; conus of the pedicel absent (Fig. 7J). Distinguished from other Empididae in the males by the presence of highly sclerotized cercus involved in clasping during mating; or reduced and membranous cercus (Fig. 5G). Female tergite 10 complete and without spine-like setae (Fig. 5I).

Empidinae Latreille
Comments. This subfamily includes the tribes Hemerodromiini and Chelipodini previously assigned to Hemerodromiinae. The genera Afrodromia and Drymodromia were removed from Chelipodini by Plant (2011a) and placed incertae sedis within Hemerodromiinae. Following our analysis these two taxa are now unplaced within the subfamily Empidinae.
Comments. This tribe is morphologically similar to Empidini, but it was recovered as sister group to Hemerodromiini in our analysis.
Diagnosis. Share similarities with those in the Dolichopodidae, particularly in rotation of genitalia and wing characters. Male terminalia rotated dextrally between 45 ∘ and 90 ∘ , excluding segment 7 (Fig. 5D), in the Dolichopodidae rotated between 90 ∘ and 180 ∘ . Wings of the Hybotidae always with a simple vein R 4+5 ; costa ends near or at M 1 /M 1+2 but may also end at or near R 4 + 5 /R 5 (Fig. 6I). Distinguished from Dolichopodidae also by the point of origin of vein R s , being at a distance from the humeral crossvein (h) equal to or longer than length of h.
Comments. The diagnosis of the tribus Bicellariini, here elevated to subfamily, and of the subfamily Hybotinae, are revised below, following the results from our analysis. Diagnosis. Similar to those in the Tachydromiinae in that the wings lack cell dm. Separated by having the M 1+2 vein branched into M 1 and M 2 (Fig. 6J); M 1 and M 2 may be faint or abbreviated at base; and the cup cell present and shorter than the bm cell, or as long as the bm cell. Proboscis oriented ventrally (Fig. 8E) while in the Tachydromiinae it is oriented posteriorly or dorsally.
Comments. The tribe Bicellariini is removed from Hybotinae and raised to a distinct hybotid subfamily based on our analysis, i.e. it is not closely related to Hybotini. Instead it forms the sister group to remaining Hybotidae and constitutes a few morphologically homogenous genera. Genera included. Bicellaria, Hoplocyrtoma Melander, and Leptocyrtoma Saigusa (Sinclair & Cumming, 2006). Hybotinae Meigen, 1820: 346. Type genus Hybos Meigen.

Hybotinae Meigen
Diagnosis. Characteristic wing characters, like the presence of a dm cell; the unbranched M 1+2 ; presence of a cup cell being as long as, or longer than the bm cell (Fig. 6I). Proboscis oriented strictly anterad (Fig. 8F).
Comments. The taxa previously included in Hybotinae as Bicellariini are here placed into a separate subfamily.
Diagnosis. Characteristic by having wings with intact dm cell; a branched M 1+2 together with a short M 1 or M 1 is absent; and the cup cell shorter than the bm cell (Fig. 6K). Proboscis oriented ventrad or slightly ventro-caudad (Fig. 8G).
Diagnosis. Similar to those of the Trichininae, particularly in the wings including the presence of dm cell; the branched M 1+2 ; the complete M 1 which reaches the wing margin; and the cup cell being shorter than bm cell (Fig. 6L). Distinguished from the Trichininae by the presence of anterad projecting and often long proboscis (except for species in Allanthalia Melander which have a scarcely visible proboscis) (Fig. 8H).
Comments. This group was previously classified as a tribe within the Ocydromiinae by Chvála (1983), but was elevated to subfamily by Sinclair & Cumming (2006). The latter classification was supported in our analysis. Oedaleinae formed a sister group to the Hybotinae.

Genera
included. Allanthalia, Anthalia Zetterstedt, Euthyneura Macquart, and Oedalea Meigen (Chvála, 1983;Sinclair & Cumming, 2006). Tachydrominae  Diagnosis. Similar to the Bicellariinae in the wings by the absence of the dm cell, but vein M 1+2 is unbranched (Fig. 6M). Unique in the family by the absence of a cup cell or the cell very small, at most reaching to half the length of bm cell; and the anal vein often faint. Proboscis oriented ventrad or ventrocaudad (Fig. 8I).

Tachydromiinae Meigen
Comments. This subfamily is distinguished by its wing characters. The taxon forms a sister group to the other Hybotidae subfamilies excluding Bicellariinae. Collin (1961) first divided the subfamily into Drapetini and Tachydromiini and Sinclair & Cumming (2006) introduced the third tribe Symballophthalmini.
Comments. This tribe was erected by Sinclair & Cumming (2006) to include the genus Symballophthalmus Becker, 1889. The tribus was found to represent the sister group to the remaining Tachydromiinae by Sinclair & Cumming (2006). It was in our analysis recovered as sister group to Tachydromiini, similar to the hypothesis by Chvála (1975). (Sinclair & Cumming, 2006).
Diagnosis. Similar to those of the Drapetini by nonbroadened wing (Fig. 6M). Distinguished from Drapetini by the presence of bare eyes (Fig. 8I) and prosternum fused with proepisternum into a precoxal bridge (Fig. 7G).
Diagnosis. Similar to those in Oedaleinae in that the dm cell is present, the M 1+2 vein branched, and the cell cup shorter than the bm cell (Fig. 6P). Distinguished from the Oedaleinae by the presence of a short proboscis pointing ventrally (Fig. 8K).
Comments. This group was removed from Ocydromiinae by Sinclair & Cumming (2006). The species in our analysis formed a monophyletic group sister to (Ocydromiinae, (Hybotinae, Oedaleinae)).

Genera included. Trichina Meigen and Trichinomyia
Tuomikoski (Chvála, 1983). Diagnosis. Similar to those in the Empididae and Atelestidae by having symmetrical and straight terminalia (Fig. 5E), as well as the point of origin of vein R s situated at a distance from the humeral crossvein (h) as long as or longer than h (Fig. 6Q). Separated from those in the Atelestidae by the presence of circumambient costa in the wing (Fig. 6Q). Distinguished from those in the Empididae by having a prosternum separated from proepisternum (Fig. 7H), except for Hydropeza spp. that can be separated by the recurved labrum, which is otherwise straight in Empididae.
Comments. Excluding Iteaphila and Anthepiscopus from Empididae, Sinclair & Cumming (2006) suggested this group to be a distinct higher taxon. It was formally published as a new subfamily within Empididae by Sinclair (2016). The group ((Iteaphila, Anthepiscopus), Ragadinae) is here recognized as forming the sister group to Empididae. Ragadinae is raised to family level, based on genetic differences separating the group from Empididae. It is expanded to include a new subfamily Iteaphilinae, comprising the two genera Iteaphila and Anthepiscopus.
Diagnosis. Distinguished from those in the Ragadinae by the absence of ventroapical sclerites (epipharyngal blades) on the labrum (Fig. 8L); absence of a ventroapical comb on the labrum; and the third segment of antenna (postpedicel) at least three times longer than wide (Fig. 7L).
Comments. This new subfamily includes two very similar genera previously placed as incertae sedis within the superfamily Empidoidea by Sinclair & Cumming (2006). It was in our analysis found to be sister group to the subfamily Ragadinae.
Diagnosis. Very similar to those of the Iteaphilinae, separated by having a labrum with ventroapical sclerites (epipharyngal blades) (Fig. 8M) and ventroapical comb; third antennal segment (postpedicel) at most two and a half times as long as broad (Fig. 7M).
Comments. This group comprises a number of genera as proposed by Sinclair & Cumming (2006) and later classified in a separate subfamily as described by Sinclair (2016).

Supporting Information
Additional supporting information may be found online in the Supporting Information section at the end of the article. Table S1. Included taxa with voucher and GenBank accession codes. Table S2. Primer names and sequences used in amplification of genes in this study. Table S3. PCR reaction protocols for each gene fragment. Table S4. Complete statistics from substitution saturation tests for codon 1+2 and 3 using Xia's method (Xia et al., 2003;Xia & Lemey, 2009) in dambe 6.4.29 (Xia, 2013).