Phylogenetic relationships of Microdontinae (Diptera: Syrphidae) based on molecular and morphological characters

The intrasubfamilial classification of Microdontinae Rondani (Diptera: Syrphidae) has been a challenge: until recently more than 300 out of more than 400 valid species names were classified in Microdon Meigen. We present phylogenetic analyses of molecular and morphological characters (both separate and combined) of Microdontinae. The morphological dataset contains 174 characters, scored for 189 taxa (9 outgroup), representing all 43 presently recognized genera and several subgenera and species groups. The molecular dataset, representing 90 ingroup species of 28 genera, comprises sequences of five partitions in total from the mitochondrial gene COI and the nuclear ribosomal genes 18S and 28S. We test the sister‐group relationship of Spheginobaccha with the other Microdontinae, attempt to elucidate phylogenetic relationships within the Microdontinae and discuss uncertainties in the classification of Microdontinae. Trees based on molecular characters alone are poorly resolved, but combined data are better resolved. Support for many deeper nodes is low, and placement of such nodes differs between parsimony and Bayesian analyses. However, Spheginobaccha is recovered as highly supported sister group in both. Both analyses agree on the early branching of Mixogaster, Schizoceratomyia, Afromicrodon and Paramicrodon. The taxonomical rank in relation to the other Syrphidae is discussed briefly. An additional analysis based on morphological characters only, including all 189 taxa, used implied weighting. A range of weighting strengths (k‐values) is applied, chosen such that values of character fit of the resulting trees are divided into regular intervals. Results of this analysis are used for discussing the phylogenetic relationships of genera unrepresented in the molecular dataset.


Introduction
The Microdontinae, a subfamily of Syrphidae (Diptera), are distributed over all continents except Antarctica. The vast majority of the 454 described species occurs in the tropics, of which 203 are in the Neotropics. With 64 species known from the Holarctic region, the group is less well represented in temperate regions (Reemer & Ståhls, 2013), perhaps explaining the limited taxonomic attention.
The most recent advocates of a family status for Microdontinae are Thompson (1972) and Speight (1987Speight ( , 2010, based on the 'basal' relationship of Microdontinae with other Syrphidae as inferred by Thompson (1969) from a Hennigian argumentation scheme of characters considered of critical importance. Speight (1987) found additional morphological differences between Microdontinae and other Syrphidae used to support family status. Recent studies have confirmed this sister-group relationship (Skevington & Yeates, 2000;Ståhls et al., 2003;Rotheray & Gilbert, 2008). Although most recent authors consider the group as a subfamily of the Syrphidae (Ståhls et al., 2003;Cheng & Thompson, 2008), others prefer family rank (Speight, 2010).
Classification of Spheginobaccha de Meijere has received special attention by several authors and its phylogenetic position has shifted between different subfamilies of Syrphidae (reviewed by Thompson, 1974). The first to include this genus in the Microdontinae was Hull (1949), after which Thompson (1969) excluded it, but Shatalkin (1975a) included it again. Based on a phylogenetic analysis of combined morphological and molecular data, Ståhls et al. (2003) placed Spheginobaccha in the Microdontinae with the genus assigned as the sister group to all other Microdontinae.
Previous phylogenetic hypotheses concerning Microdontinae relied on too few taxa in relation to the high morphological diversity, for example, two in Skevington & Yeates (2000) and six in Ståhls et al. (2003) and Hippa & Ståhls (2005). Relationships within the Microdontinae remain completely unaddressed. Specifically, the supposed sister-group relationship of Spheginobaccha and the other Microdontinae needs testing by much broader taxon sampling. For instance, Aristosyrphus Curran, Eurypterosyrphus Barretto & Lane and Mixogaster Macquart share features with Spheginobaccha, such as a hypandrium with apical part consisting of two separate lobes, an unfurcate phallus and characters of wing venation (Reemer & Ståhls, 2013). An extended taxon set maximally representing genus groups (whether recognized previously or not) better tests this sister-group relationship.
Here we examine the phylogenetic relationships of the (sub)genera within the subfamily, based on molecular and morphological characters. We analyse a combination of morphological and molecular characters from a large set of microdontine taxa to test the sister-group relationship of Spheginobaccha with the other Microdontinae, attempt to elucidate phylogenetic relationships within the Microdontinae and discuss uncertainties in the classification of Microdontinae as proposed by Reemer & Ståhls (2013). The rank of the group in relation to the other Syrphidae is discussed briefly.

Selection of ingroup taxa and specimens
The starting point for the selection of ingroup taxa for morphological analysis were the genus group names of Microdontinae listed by Cheng & Thompson (2008), and further revised by Reemer & Ståhls (2013). At least one species, preferably the type species, of all groups was included in the morphological dataset. Exceptions are objective or otherwise obvious synonyms (e.g. Aphritis Macquart, Colacis Gistel, Holmbergia Lynch Arribalzaga) and taxa based only on immature stages (e.g. Ceratoconcha Simroth, Nothomicrodon Wheeler) (for more information on these names and synonymies, see Cheng & Thompson, 2008). Many new or little known species were included which had been unassigned previously, or were lumped under Microdon despite morphological peculiarities. Descriptions can be found in Reemer & Ståhls (2013). When possible, more than one species per genus group was included: for Spheginobaccha the six species represent all three species groups of Thompson (1974).
The studied specimens were obtained from many sources. When possible, the primary types were studied, especially when no additional material was available. All specimens used for constructing the morphological matrix are listed in Table S1.
Characters were scored mostly from males. For four taxa of which only females were available, male genital characters were derived from closely related species. This approach was implemented only for taxa closely similar in external morphology, and which obviously belong to the same genus or species group (indicated in the 'remarks' column in Table S1).
In the molecular dataset, as many taxa as possible were included, depending on the availability of specimens for molecular analyses. The list of specimens used for DNA extraction, including locality and collection data as well as GenBank accession numbers, is given in Table S1. Morphological and molecular characters are based on specimens of the same species, except for Rhopalosyrphus ramulorum Weems & Deyrup for which morphological characters are based on a specimen of the closely related R. guntheri (Lynch Arribalzaga).
Specimens used for DNA extraction came from a wide variety of sources and collection methods. Fresh material (up to 3 years old) collected directly into ethanol was rare, so older material (up to about 10 years old), sometimes preserved dry, was used also. PCR results thus differed strongly among the taxa and among the genetic markers.

Morphological character matrix
The starting point for the morphological character matrix was Hippa & Ståhls (2005); many characters from this proved useful. Others were modified (e.g. extra character states were added), and some omitted because of irrelevance or for pragmatic reasons (see below). The numbering system of Hippa & Ståhls (2005) is indicated by the letters H&S, to avoid confusion with those in this matrix.

Notation of character statements
Following Sereno (2007), we distinguish between characters (as independent variables) and character states (as mutually exclusive conditions of a character). The description of a character is subdivided hierarchically into a secondary locator L2 (e.g. antenna), a primary locator L1 (e.g. basoflagellomere), a variable V (e.g. length) and a variable qualifier q (e.g. length relative to scape). The states are given subsequently following a colon. A secondary (or even tertiary) locator is added only to clarify the position of the primary locator. In the example given above, the entire character statement could be as follows: Antenna, basoflagellomere, length relative to scape: shorter (0); as long as (1); longer than (2).

Drawings and photographs
Male genitalia were dissected and macerated in an aqueous 10% KOH solution at ambient temperature for 12-24 h and stored in glycerol, in microvials attached to the original specimen. Drawings of male genitalia were made with the aid of a drawing tube attached to a Wild M20 compound microscope. Digital photographs of (parts of) specimens were taken through an Olympus SZX12™ motorized stereozoom microscope, using Analysis Extended Focal Imaging Software.

Choice of molecular markers
For the molecular dataset, five sequence partitions of three molecular markers were used: two each for the mitochondrial COI gene and the nuclear ribosomal RNA gene 18S , and one for the nuclear ribosomal RNA gene 28S . Primer information and combinations are given below and in Table S2. Molecular markers derived from previous studies on Syrphidae -a combination of mitochondrial COI and nuclear 28S sequences with morphological characters -yielded good results in the study on the intrafamilial relationships of Syrphidae of Ståhls et al. (2003). The 18S gene fragment used by Mengual et al. (2008) proved informative for reconstructing deeper branches in the study of relationships within the subfamily Syrphinae.

DNA extraction
For most specimens, two or three legs were used for DNA extraction. In a few cases the entire thorax or the abdomen was used. Prior to extractions, ethanol-preserved samples were rinsed in distilled water.
DNA extractions were made using the NucleoSpin ® Tissue extraction kit, following the manufacturer's protocol, eluting the DNA into 50 μL of elution buffer. For some very small specimens NucleoSpin ® Tissue XS was used, which involves the same extraction procedures, except for some differences in the quantities of buffers and washing liquids.
The PCR products were visualized by running 4-μL PCR products on a 1.5% agarose gel. PCR products were treated with ExoSapIt prior to sequencing reactions. Sequencing electrophoresis was carried out in the sequencing service laboratory of the Finnish Institute for Molecular Medicine (FIMM), University of Helsinki, Finland, with an ABI3730xl DNA Analyzer.
Sequences of forward and reverse primers were assembled and edited in Sequence Navigator v1.01 (Applied Biosystems, Foster City, CA). For the outgroup taxon Chalarus spurius (MZH-voucher Y0800), the COIb sequence was not available, for which reason the sequences of this taxon were combined with the COIb sequence of Chalarus sp. (MZH-voucher Y0038).

Alignment
The mitochondrial DNA sequences of the (protein coding) COIa and COIb gene fragments were aligned manually by their codon positions. Sequences of the 18S and 28S ribosomal RNA genes were aligned separately using MAFFT v6 (Katoh et al., 2002(Katoh et al., , 2009Katoh & Toh, 2008). This program offers several algorithms, some of which perform very well compared to those of other programs (e.g. ClustalW, DIALIGN-T, T-COFFEE) for multiple sequence alignment (Golubchik et al., 2007;Rosenberg, 2009). We used E-INS-i: based on Katoh & Toh (2008) and Katoh et al. (2009), this algorithm was considered most suitable for ribosomal DNA sequences as it was developed for dealing with sequences with considerable length variation.

Datasets and analyses
Phylogenetic analyses were made by both parsimony and Bayesian analyses.
Parsimony analyses were performed using the software program TNT (Tree Analysis using New Technologies) v1.1, October 2010 (Goloboff et al., 2008b), with gaps treated as missing data and morphological characters treated as nonadditive. Two sets of taxa were used (with the same set of morphological characters): a set containing all 189 taxa (the 'total set') and a set containing only the 90 taxa for which DNA data are available (the 'subset').
All molecular markers were analysed separately and sequences with remarkable placements (e.g. ingroup taxa in the outgroup) were scrutinized for possible errors in the sequences, such as copy-paste errors in the datafiles or contamination during DNA extraction or amplification. A few suspect or erroneous sequences were omitted subsequently from further analyses.
A single matrix integrating the data of all three molecular markers (in five fragments) then was constructed, which contained 90 taxa and 2740 columns of nucleotide data (including gaps). The TNT search for this matrix was stopped after the shortest length was found 50 times. Molecular and morphological datasets were merged using the dmerge command in TNT, resulting in a concatenated matrix of 2740 molecular sites and 174 morphological characters.
Combined matrices were analysed using a combination of all four 'new technology' heuristic search methods of TNT, under their default parameters: sectorial search, parsimony ratchet, tree-drifting and tree-fusing (see, e.g., Goloboff et al., 2008b for explanations on commands). Searches were set to stop when minimum tree length was hit 100 times for the subset and 30 times for the total set.
Morphological characters of the total set of taxa were subjected to parsimony analyses under 'implied weighting' (Goloboff, 1993;Goloboff et al., 2008a), a method which downweights characters according to their degree of homoplasy. The strength of the weighting function is determined by constant value k in the implied weighting formula. The approach used here derives from Mirande (2009), who explored a range of k -values. In this approach, the chosen kvalues were not distributed regularly, because -as Mirande (2009) argues -this results in an artificial bias of the results towards the higher k -values. This bias is avoided by choosing k -values such that the values of fit (F ) produced by the trees obtained under different k -values are divided into regular intervals. Here, as in Mirande (2009), k -values were chosen so as to result in average character fits of 50, 54, 58, 62, 66, 70, 74, 78, 82, 86 and 90%. In order to obtain k -values, the formula for implied weighting was rewritten as [k = (F × S )/(1-F )]. S is a measure of the average homoplasy per character, calculated as [S = (number of observed steps) -(minimum number of steps)/(number of characters)]. The number of observed steps is based on the shortest trees found under equal weights (2292 for the total set of 189 taxa). The minimum number of steps is the cumulative number of minimum character state changes for all 174 characters, which amounts to 242. So, the value of S used for the total set of taxa is (2292 to 242)/174 = 11.78. The resulting k -values are listed in Table S3. As in Mirande (2009), the most stable trees are considered to be those four which share the highest number of nodes with the other trees, as measured by the SPR-distance (Goloboff, 2008) and the distortion coefficient sensu Goloboff et al. (2008b) (DCG), which was determined using the tcomp command in TNT. A strict consensus tree was derived from these four trees.
For Bayesian analyses, MrBayes v3.2 was used (Huelsenbeck & Ronquist, 2001;Ronquist & Huelsenbeck, 2003). Molecular data were divided into nine partitions: 18S 1, 18S 2, 28S , and three for each codon position of CO1 a and CO1 b. The general time reversible nucleotide substitution model with invariant gamma (GTR + I + G) was applied to these sequences, with separately calculated sets of parameters for each partition. The morphological dataset was analysed under the mk1 model (Lewis, 2001), with coding set to variable. Each analysis consisted of two independent simulations of four simultaneous MCMC chains, sampling every 1000th generation. Convergence was reached after 20 million generations with 0.038 (molecular data only) and 0.014 (molecular and morphological data combined) standard deviation of split frequencies, as suggested by Ronquist et al. (2011). The initial 5000 trees (25%) were discarded as burn-in. Majority-rule consensus trees were computed with posterior probabilities for each node.

Measures of support and stability
Bremer support values were calculated by TBR branch swapping based on the strict consensus trees, using the 'Bremer supports' option under the 'Trees' menu, examining trees up to 100 steps longer than the most parsimonious trees. Jackknife values and GC frequency differences (Goloboff et al., 2003) were calculated in TNT, using 1000 replicates and a removal probability of 36%. GC values indicate the difference between the frequency in which nodes are retrieved in the jackknife replicates and the frequency of the most frequent contradictory group. So, in contrast to normal jackknife values, the GC values are informative for the amount of contradictory information in the dataset. If these values are equal, there are no contradictory groups that are supported by the data. For Bayesian inference, posterior probability values are indicated.
In Hippa & Ståhls (2005) (# 44), pilosity of the katepisternum is coded into one character statement. In Microdontinae, the katepisternum is never entirely pilose: the dorsal and ventral patches of pile are always widely separated. The dorsal pilosity always is close to the dorsal margin, whereas ventral pilosity is mostly very sparse and only found close to the ventral margin. Presence of pile on the dorsal part is here considered to be independent of presence of pile on the ventral part, and therefore are coded in separate statements (056 and 057). 057. Katepisternum, ventral part, pilosity: bare (0); pilose (1).
In most Microdontinae, the abdominal spiracle in the metepimeron is surrounded by a dense fringe of long microtrichia, often forming a sort of tuft. In a few taxa this fringe is absent. 073. Mesonotum, transverse suture, presence: absent or only weakly visible at notopleuron (0); well-developed, but incomplete and may be short (1) (1); absent (2). H&S # 052.
As there is no straightforward division between the two states, the coding of this character is quite subjective. Although in many taxa the pile/setae under consideration are thicker than on other parts of the femur, state 1 was chosen only for few taxa. 119. Femora, ventral surface, pilosity: entirely pilose (0); with bare median stripe limited to apical half (1); with bare median stripe extended to basal half (2). H&S # 062. Hippa & Ståhls (2005) (Fig. 52); present (1) (Fig. 53).
In several (mainly Neotropical) taxa the hind tibia is occupied with long, dense pile, reminescent of the corbicula of bees. In these taxa the hind tibia is often also strongly widened, which adds to the resemblance to bees. 126. Mid tarsus, basitarsomere, ventral vestiture: without spine-like setae (0), with pale spine-like setae (1), with dark spine-like setae (2). H&S # 067. 127. Hind basitarsus of male, dorsal view, width: as wide as (0); wider than (1) apex of hind tibia. This character is often sexually dimorphic: often state 1 is most pronounced in the male and less so or even absent in the female.
The presence of pilosity on sternite 1 seems to be of good diagnostic value for certain genera or species groups, as little variation was found in this character among closely related species. 144. Sternite 2, anterior sclerite, presence: absent (0); present (1) (Figs 63, 64).
In several taxa, the hypandrium is depressed laterally, and the lateral depressions are delimited ventrally by a sharp ridge. The ridge may be very close to hypandrial margins and may be overlooked. Table S1 indicates which fragments could be amplified for each sample. Total success rates for the different fragments were as follows: COIa -(89%); COIb (84%); 18Sa (99%); 18Sb (69%); 28S (69%).

Phylogenetic analyses
Analyses of the the molecular dataset under parsimony resulted in 22 most parsimonious trees of length 8662 (strict consensus in Figure S1) and under Bayesian gave the tree in Figure S2.
Analyses of combined molecular and morphological datasets under parsimony resulted in three most parsimonious trees of length 9965 (strict consensus in Fig. 82) and under Bayesian gave the tree in Fig. 83.
Parsimony search of the morphological dataset for the total set of 189 taxa yielded 11 trees for the different k -values, four of which were used to construct a strict consensus ( Figure S3).

Evaluation of trees
The two trees based on the analyses of molecular data alone (strict consensus parsimony and Bayesian) are poorly resolved (Figures S1, S2). Most taxa are resolved within a large polytomy of Microdontinae, in which only a few small clades are recovered. In both trees, the Microdontinae are placed as sister group of other Syrphidae, and Spheginobaccha is placed as sister to the other Microdontinae, with high support. Otherwise, there are few corresponding nodes and most support values are low.
The addition of morphological characters to the dataset clearly adds much resolution to the trees (Figs 82, 83). Support remains generally low for the deeper nodes, but many more derived nodes have higher values. Again, Microdontinae are recovered as sister group of other Syrphidae, and Spheginobaccha is placed as sister to all other Microdontinae.
Following the reasoning of Kluge (1989) concerning the philosophy of total evidence in phylogenetic analyses, the results obtained from a combination of morphological and molecular data are preferred to those from a single data source. Therefore, the trees based on combined analyses (Figs 82,83) were chosen as preferred results. However, several genus groups are not represented in combined analyses because of absence of molecular data. Thus, an analysis of morphology alone is presented here. This analysis includes 189 taxa, representing all genera and species groups as recognized by Reemer & Ståhls (2013) (Figure S3). The phylogenetic position of taxa which could not be included in combined analyses will be discussed based on these results.

Family affairs
Our results support the sister-group relationship of Microdontinae and other Syrphidae, as proposed originally by Thompson (1969) and subsequently by others (Skevington & Yeates, 2000;Ståhls et al., 2003;Hippa & Ståhls, 2005;Rotheray & Gilbert, 2008). Our results are based on a wide representation of taxa: representatives of all valid genus groups are included, as well as taxa from all major biogeographic regions. In addition, both character sets (molecular and morphological) are larger than in previous analyses and the results can be regarded as additional support for this sister-group relationship. However, the results cannot be regarded as compelling evidence as analyses were not designed to test this relationship explicitly. For that, a much larger set of Syrphidae taxa would be necessary and preferably more taxa of related groups of 'lower Cyclorrhapha' included, such as Phoridae and Platypezidae.
According to Speight (2010) the presumed sister-group relationship between Microdontinae and other Syrphidae 'more-or-less reduces the issue of the correct placement of Microdon and allied genera to a matter of personal preference'. We advocate, however, that in this case, in which available evidence does not demand that the classification be changed, a conservative attitude is preferable.

Family group names
Two tribes are recognized within the Microdontinae: Spheginobacchini Thompson, 1972, which includes only the genus Spheginobaccha, and Microdontini Rondani, 1845 including all remaining taxa (Cheng & Thompson, 2008). The only other proposed family group names are Masarygidae of Brèthes (1908) and Ceratophyini of Hull (1949), unused since their introduction. Hull (1949) wrote: 'Perhaps two tribes should be recognized. The first would be the Microdonini distinguished by ( . . . ), and secondly the Ceratophyani ( . . . ). ' Sabrosky (1999) argued that this name is unavailable, as it was mentioned only casually within a short diagnosis of a group, not as a formal proposal of a new group name. However, this can be regarded as a 'conditional proposal' of a new name: as published before 1961, there is no reason for considering this name unavailable (ICZN, 1999: art. 15.1).
Recognition of additional tribes could make the group more 'manageable' in taxonomic, biogeographical and evolutionary studies and discussions. However, for introducing new family group names (or changing the status of available ones), we feel that the clades under consideration should be sufficiently 'reliable'. Our support values (Figs 84, 85) aid in assessing clade reliability. For most larger clades, these values are low, but smaller clades for which these values are higher, are here -subjectively -considered to be of generic level, rather than of family-group level. Because of this, and also because of the considerations on missing data as discussed in the previous paragraph, the introduction of new tribal names or reinstating available family group names based on the present phylogenetic hypotheses is deemed unjustified.

Intra-and infrageneric relationships
Several smaller clades have high support and stability values, and indicate affinities between genus and species groups that have not been suggested previously. We address these affinities, and discuss the classification of Reemer & Ståhls (2013).
Afromicrodon. This genus is recovered at an early node in all trees, but at varying positions. However, support values are quite low, so the position is not clarified.
Archimicrodon. In combined analyses, two of the three groups recognized by Reemer & Ståhls (2013) are represented: Hovamicrodon (unidentified species) and Archimicrodon s.l. (clatratus and simplex ). These taxa were recovered as a clade. The analysis based on morphological characters also includes two species of Archimicrodon s.s.: A. malukensis sp.n. and A. simplicicornis. These species are united in a clade within a large polytomous clade, which offers no hypothesis as to the relationships with Archimicrodon s.l. and Hovamicrodon.
The three groups are very similar in morphology, and are likely to be closely related. The subgenus Hovamicrodon probably is monophyletic, considering the spatulate scutellar calcars and distribution restricted to Madagascar. However, as the phylogenetic results indicate, it is so closely related to Archimicrodon s.l. (which is recovered as paraphyletic with respect to Hovamicrodon) that separate generic status seems unwarranted. Besides, a spatulate shape of the scutellar calcars can be found also in certain species of the New World groups Laetodon and Serichlamys. The latter genus is recovered as sister to Archimicrodon in combined analyses. As this  character is not unique, it provides insufficient basis on which to base a genus.
Aristosyrphus. No molecular data available. In morphological analyses the subgenera Aristosyrphus and Eurypterosyrphus are recovered as sister groups. This does not contradict the present rank of Eurypterosyrphus as subgenus of Aristosyrphus (Cheng and Thompson, 2008;Reemer & Ståhls, 2013). Considering the large morphological variation within this genus, especially within the subgenus Eurypterosyrphus, both in external characters and male genitalia, the phylogenetic relationships of these taxa need to be examined in more detail, preferably with the aid of molecular characters.
Although Aristosyrphus and Mixogaster were not recovered as closely related groups, certain morphological characters in common to these taxa may suggest a closer relationship. For instance, in some specimens of Aristosyrphus primus an anterior stump is present at vein M (Fig. 48). This character always has been diagnostic for Mixogaster (Hull, 1954;Cheng & Thompson, 2008). A facial tubercle similar to that of Eurypterosyrphus is present also in certain species of Mixogaster. In addition, the genera share an unfurcate   Bardistopus. No molecular data available. In morphological analysis Bardistopus is placed as sister to a clade containing several taxa in which the males have a bifurcate basoflagellomere: Schizoceratomyia, Furcantenna and Carreramyia. In Bardistopus the basoflagellomere is not furcate. Tentatively, a placement with Paramixogaster seems more plausible, because these taxa share the following characters: basoflagellomere much longer than scape, not furcate; postpronotum bare; vein R4 + 5 with posterior appendix; phallus strongly bent dorsad, relatively deeply furcate. Unlike Paramixogaster the abdomen is not constricted in dorsal view, but tergite 2 in lateral view clearly is flattened relative to tergites 3 and 4.
Carreramyia. Carreramyia megacephalus is one of the microdontine taxa in which the basoflagellomere of the male is bifurcate. When Shannon (1925) described this species, he attributed it to Microdon, denying the furcate antenna warranted erection of a new genus, as this condition is restricted to the male. van Doesburg (1966) did not agree and considered Microdon megacephalus to be very different from other Neotropical taxa with furcate basoflagellomeres (Masarygus and Schizoceratomyia), and hence erected for it the genus Carreramyia. Cheng & Thompson (2008) considered Carreramyia megacephalus as a Ubristes species with furcate basoflagellomere, a character they considered   (Reemer & Ståhls, 2013), we do not propose synonymy.
Ceratophya. In combined analyses, Ceratophya sp. was placed within Stipomorpha, as follows: [(C . sp. nov. + S. lanei ) + (other Stipomorpha species)]. However, there are several important morphological differences between Ceratophya and Stipomorpha (e.g. tergites 3-4 fused vs. not fused, sternites 2-3 widely separated vs. narrowly separated, phallus furcate vs. unfurcate). It seems wise to wait with making changes in the taxonomy of these genera until more species can be included in the molecular dataset.
Ceriomicrodon. Based on morphology alone, this taxon was placed in the clade also comprising Domodon, Pseudomicrodon and Rhopalosyrphus. In male genitalia, Ceriomicrodon is very similar to these taxa, having in common a strongly elongate, whip-like dorsal process of the phallus. It resembles Rhopalosyrphus in the ventrally bulging face, the antennal fossa being wider than high, the narrow area of enlarged ommatidia on the eye, and the constricted abdomen. The bare postpronotum and bare katepimeron distinguish Ceriomicrodon from Rhopalosyrphus, whereas the bare postpronotum and the flat vertex distinguish it from Pseudomicrodon.
Cervicorniphora. Morphological analysis provided few clues as to the taxonomic affinities of this taxon, although it seems unrelated to other taxa in which the male has a furcate basoflagellomere.
Chrysidimyia. Morphological analysis placed Chrysidimyia in a large polytomy, leaving its phylogenetic affinities unresolved. As Reemer & Ståhls (2013) argue, the male genitalia of Chrysidimyia resemble those of Laetodon; these taxa share an unfurcate phallus and a long posterior process on the phallus. These taxa also have in common metallic body coloration and pilose eyes.
Chymophila (subgenus of Microdon). Combined analyses included one Oriental and one Neotropical species, which are recovered as sister species within Microdon. Dimeraspis (subgenus of Microdon). Morphological analysis includes three species belonging to this taxon (M. abditus, M. fuscipennis, M. globosus), but the results offer little clues as to their relationships. Because of similarities in male genitalia this group might be related to Archimicrodon, Menidon or Serichlamys, but is continued to be treated as subgenus of Microdon.
Domodon. No molecular data are available. Morphological analysis placed Domodon in a clade with Ceriomicrodon, Omegasyrphus, Pseudomicrodon and Rhopalosyrphus. These genera have in common a strongly elongate, whip-like dorsal process of the phallus.
Furcantenna. Based on morphology, Furcantenna nepalensis was recovered in a clade containing Carreramyia and Schizoceratomyia. Furcantenna is very similar to Schizoceratomyia in both external morphology and male genitalia, but presently available evidence is not conclusive about the exact relationships between these taxa.
Heliodon. Five species of Heliodon are included in combined analyses. These are recovered in a clade containing also Indascia; thus, Heliodon appears as paraphyletic with respect to that genus. However, support for the subclade containing the Indascia species is low. As Heliodon morphology is distinct from that of Indascia (Reemer & Ståhls, 2013), these taxa will continue to be considered as separate.
Hypselosyrphus. In combined analyses (Figs 82, 83), Hypselosyrphus was recovered in a clade together with Rhoga, with high support. Hypselosyrphus is paraphyletic with respect to Rhoga, as seen in morphological analyses. However, morphological variation within Hypselosyrphus is large, which could indicate a more complicated phylogeny so Hypselosyrphus and Rhoga are retained as separate genera.
Indascia. Three species of Indascia are included in combined analyses: these are recovered in a well-supported clade, part of a larger clade containing also Heliodon, which appears paraphyletic with respect to Indascia. However, support values are low, and as Indascia is distinct from Heliodon in morphology, these taxa are considered separate genera.
Superficially, species of Indascia look similar to those of Paramicrodon (as noticed by Cheng & Thompson, 2008), but available phylogenetic evidence provides no support for a close relationship. See Paramixogaster for further discussion.
Kryptopyga. Morphology provides no evidence for a close relationship with Ptilobactrum; Kryptopyga pendulosa is placed as sister of Ceratrichomyia. These taxa share the pilose basoflagellomere in the male, the swollen vertex and dorsal occiput, and the unfused tergites 3 and 4. Male genitalia are quite different, and in Kryptopyga the mesonotal transverse suture is incomplete.
Laetodon. The included species (Reemer & Ståhls, 2013) were placed previously in Microdon (Thompson, 1981). In combined analyses, Laetodon geijskesi (Doesburg) is placed quite distant from Microdon. The analysis of morphology alone includes an additional species, L. laetus (Loew), but provides no alternative hypothesis as to the relationship with Microdon. See Chrysidimyia for further discussion.
Masarygus. Because no fresh specimens of the type species of Masarygus, M. planifrons, were available, M. palmipalpus Reemer was substituted. Morphological analysis placed the two species as sister taxa, but in combined analysis, M. palmipalpus was placed as sister of Carreramyia tigrina, with moderate support. The clade including both taxa was placed as sister of (Paragodon + Surimyia), with low support.
Two undescribed species belonging to this genus are included in morphological analysis under Masarygus sp. 1 and sp 2 (Reemer & Ståhls, 2013). Whereas sp. 1 is placed in the same clade as M. planifrons and M. palmipalpus, the relationships of sp. 2 are unresolved.
Megodon (subgenus of Microdon). Megodon stuckenbergi was included in the analysis of morphological characters, which recovered it within a clade also containing Microdon s.s. Exact relationships, however, remain unclear.
Menidon. Combined analysis places Menidon falcatus in a clade with Paramicrodon and Piruwa, but support values for this clade are low. Neither Bayesian analysis nor parsimony analysis based on morphology alone offers an alternative solution.
Metadon. In combined analyses, the included species of Metadon are grouped in a clade with high support. Relationships with other genera are less clear: in combined parsimony analysis, Metadon is recovered as sister of Parocyptamus, within a clade containing also Microdon s.s., Stipomorpha and Ceratophya. Bayesian analysis places it in a polytomy with Laetodon, Microdon, Omegasyrphus, Parocyptamus, Pseudomicrodon and Rhopalosyrphus. The analysis of morphology alone includes a larger number of species (also from Africa), and Metadon is placed as sister group of Heliodon, within a clade containing Ceriomicrodon, Domodon, Omegasyrphus, Peradon, Pseudomicrodon and Rhopalosyrphus.
Microdon. Microdon has served as a 'waste basket' for taxa of which taxonomic affinities were inadequate for location elsewhere. Although several taxa were assigned to other genera, subsequent authors have considered those genera as subgenera of Microdon. The present analyses contain many species formerly placed in Microdon. As can be seen in Fig. 82 (taxa classified previously in Microdon, or representatives of these taxa, are indicated with an 'M') this group is polyphyletic and its representatives are scattered over different parts of the tree. Although the exact positions of these groups may change in future analyses when more taxa and more molecular data are included, these results provide sufficient basis for subdividing Microdon into different monophyletic units (Reemer & Ståhls, 2013). For several species which could be included only in the analysis of morphological characters, however, phylogenetic affinities remain obscure ( Figure S3). These taxa will be maintained in Microdon s.l. until better phylogenetic hypotheses are available.
Mixogaster. According to combined analysis and morphology alone, Mixogaster is the first branch within the tribe Microdontini, with high support. Bayesian analysis recovers (Mixogaster + Schizoceratomyia) as the first branch within the Microdontini. Analyses of molecular data alone recover Mixogaster in shallower positions. Interestingly, the most important diagnostic character of Mixogaster, the anterior appendix of vein M, is found also in Spheginobaccha and certain specimens of Aristosyrphus primus. These taxa also share the character of the apical part of the hypandrium consisting of two separate lobes. No close relationship between Mixogaster and Aristosyrphus was recovered by the analysis of morphological characters ( Figure S3), but see Aristosyrphus (q.v.).  (Reemer & Ståhls, 2013). These are shared with Ceriomicrodon and Domodon. Even though the results of the analyses only partly support the monophyly of a clade containing these taxa, the characters of the male genitalia suggest they are related.

Myiacerapis (subgenus of Microdon
Paragodon. Thompson (1969) in describing this genus, stated that it appeared to be the 'most primitive' microdontine fly known. In both analyses of combined molecular and morphological characters, Paragodon is not placed in such a position, albeit at a relatively early node, with low support in parsimony analysis. Additional sampling of molecular characters of other taxa in the basal part of the tree will be needed. Paragodon was recovered as sister to Surimyia.
Paramicrodon. In all analyses presented here, except parsimony molecular analysis, the Neotropical Paramicrodon cf. flukei and the Oriental P . aff. nigripennis were placed together in a well-supported clade. In the analysis based on morphology three additional species (two Neotropical, one Oriental) are also resolved in a clade with the other species. No doubt the Neotropical and Oriental species belong in the same genus. Further relationships remain uncertain, and the phylogenies presented here are contradictory.
Paramixogaster. Three species of this genus included in the analyses of combined molecular and morphological characters are recovered as a monophyletic. A larger number of species was included in morphological analysis. The resulting phylogeny supports the inclusion of the following Afrotropical species in this genus, which considered previously to be Oriental and Australian in its distribution: Microdon acantholepidis Speiser, Microdon crematogastri Speiser, Microdon illucens Bezzi, Pseudomicrodon elisabethae Keiser. Paramixogaster sp. from Madagascar is also added to this genus. Morphological analysis also recovered Ptilobactrum (q.v.) within Paramixogaster.
Parocyptamus. Parocyptamus is recovered in contradictory positions: combined parsimony analysis placed it as sister group to Metadon, whereas Bayesian analysis placed it within Microdon s.l. Parsimony analysis of morphology places it in a clade with three species of Microdon s.l.
Most assigned species were included in Microdon in the most recent classification of Neotropical Microdontinae (Thompson et al., 1976), but here this group is not recovered as part of or sister to Microdon.
Piruwa. In combined analysis, this taxon was placed as sister to Paramicrodon with low support. Bayesian analysis of combined characters places it as sister to the clade containing Ceratophya and Stipomorpha.
Pseudomicrodon. The two species of Pseudomicrodon included in combined analyses are placed together in both analyses, but as sister to Omegasyrphus pallipennis in Bayesian analysis and as sister to Laetodon geijskesi in parsimony analysis (low support). In morphological analyses, Pseudomicrodon species are placed in a clade with Ceriomicrodon, Domodon, Omegasyrphus and Rhopalosyrphus, and Pseudomicrodon does not appear as monophyletic. Phylogenetic affinities between these taxa are likely because of strong similarities in male genitalic morphology (phallus with dorsal process strongly elongated). At present, the morphological basis for distinguishing Ceriomicrodon, Pseudomicrodon and Rhopalosyrphus is narrow. The groups are most probably related, but it is doubtful whether they are monophyletic, considering morphological variation.
Ptilobactrum. In morphological analysis, Ptilobactrum is placed within Paramixogaster but differences are considered too large to alter the rank of Ptilobactrum to subgenus within Paramixogaster. For instance, in contrast with Paramixogaster, the basoflagellomere and postpronotum are pilose and the abdomen is oval. Phylogenetic affinities of Ptilobactrum can best be re-assessed when molecular data become available.
Rhoga. In the molecular and combined analyses, Rhoga is recovered within Hypselosyrphus, with high support, a result found also in the analysis based on morphology only, in which more species were included. Hypselosyrphus (q.v.) is paraphyletic with respect to Rhoga.
Rhopalosyrphus. In parsimony analyses (both of molecular and combined), the two included species of Rhopalosyrphus were placed in different clades. However, in both Bayesian analyses these were placed as sisters. Analysis of morphology includes five species, four of which are placed in a monophyletic clade. In morphological analysis, Rhopalosyrphus is recovered in a clade with Ceriomicrodon, Domodon, Pseudomicrodon and Omegasyrphus. Close affinities between these taxa are likely because of similar male genitalia (phallus with strongly elongated dorsal process).

Schizoceratomyia.
Combined analyses included only one species, recovered in deep positions, but with different sister groups: Afromicrodon, Mixogaster or Paramicrodon. In analysis of morphology alone, it is placed in a clade also containing several other taxa with a furcate basoflagellomere in the male (Carreramyia, Furcantenna, Masarygus). As this grouping is lost when molecular characters are added, it appears that Schizoceratomyia cannot be treated as synonymous with Masarygus, as Hull (1949) and Papavero (1962) proposed. Johnsoniodon malleri Curran is placed within Schizoceratomyia, supporting its inclusion by Cheng & Thompson (2008) and Reemer & Ståhls (2013).
Serichlamys. Serichlamys differs from Microdon s.s. most in genital features: phallus furcate apically, hypandrium with bulb-like base, surstylus with long, ventrally directed lobe (Reemer & Ståhls, 2013). Its independent phylogenetic position from Microdon s.s. is confirmed by its sistergroup relationship with Archimicrodon. The type species of Serichlamys (S. rufipes) and S. scutifer are not recovered with other Serichlamys in morphology analysis (but in a large polytomy), but the similarities in wing venation and male genitalia are considered great enough to classify all considered species in this genus.
Spheginobaccha. Hull (1949) was the first to include Spheginobaccha in the Microdontinae. Thompson (1969) excluded it, after which Ståhls et al. (2003) included it again based on a sister-group relationship of Spheginobaccha to all other Microdontinae, as recovered in combined analyses. All our analyses affirm Spheginobaccha as a sister to all other Microdontinae, with high support. Thompson (1974) recognized three species groups: the Oriental macropoda group (Spheginobaccha s.s. in Cheng & Thompson, 2008), the African rotundiceps group (subgenus Dexiosyrphus) and the African perialla-group. Our results suggest that the perialla group (represented by S. guttula in the dataset) is sister to rotundiceps group + macropoda group, as noted by Thompson (1974).
Stipomorpha. Combined phylogenetic hypotheses placed Ceratophya sp. within Stipomorpha, as did Bayesian analysis of molecular data. Analysis of morphology alone, which includes additional species of Ceratophya and Stipomorpha, recovers both as monophyletic. Given several important morphological differences between Stipomorpha and Ceratophya (e.g. tergites 3-4 fused vs. not fused, sternites 2-3 widely separated or narrowly separated, phallus furcate or unfurcate), conclusions about the taxonomic status of these genera should await wider molecular data.
Sulcodon. No molecular data available. Morphology provides no clues on the affinities of this taxon, as it was placed in a large polytomy containing other species of Microdon as well as species of several other genera.
Surimyia. When Surimyia was described, a species assigned previously to Paragodon was included (P. minutula Doesburg). Several morphological characters indicate differences between these genera (Reemer, 2008). In both combined analyses, and also in Bayesian analysis of molecular data, Paragodon and Surimyia are recovered with low support as sister groups. However, fundamental morphological differences between the taxa (Reemer & Ståhls, 2013) suggest that, although closely related, they should be considered as separate genera.
Syrphipogon (subgenus of Microdon). Hull (1937) erected Syrphipogon, mentioning that it is related to Microdon. Steyskal (1953) referred to Hull's description in his own description of an apparently very similar species, but considered the differences from Microdon insufficient for generic status. Morphological analysis places Syrphipogon fucatissimus in an unresolved clade which also contains Microdon s.s., but provides no clues as to their relationships. In external characters and male genitalia these taxa are quite similar, and thus it seems justified to treat Syrphipogon as a subgenus of Microdon, as proposed by Reemer & Ståhls (2013).
Thompsodon. No molecular data available: known from one female specimen only, so the male genitalia are unknown. As these characters are important in morphological analysis, this genus was excluded from analysis.
Ubristes. Thompson et al. (1976) included in Ubristes the type species of Carreramyia, Hypselosyrphus and Stipomorpha. The latter two groups were considered as 'subgroups' of Ubristes by Cheng & Thompson (2008). Based on morphology alone, Ubristes flavitibia was placed in an unresolved clade with (amongst others) Microdon s.s., but to the exclusion of Carreramyia, Hypselosyrphus and Stipomorpha, supporting Reemer & Ståhls (2013) treatment of these taxa as separate genera.
Undescribed genus #2. No molecular data available: for morphology see Reemer & Ståhls (2013). In morphological analysis, this taxon is placed in a clade which contains other genera with a furcate male basoflagellomere (Carreramyia, Furcantenna, Masarygus, Schizoceratomyia). However, these genera are not recovered in a clade in combined analyses, so whether 'Undescribed genus #2' relates to any remains to be seen.

Concluding remarks
Our phylogenetic trees contain many poorly supported nodes and contradict each other on many points. Obviously, this is not the 'final word' concerning phylogenetic relationships of Microdontinae taxa. In our molecular dataset, only 28 of the 43 genera recognized by Reemer & Ståhls (2013) could be included, and more should be sought. More resolution can be expected also with increased taxon sampling within heterogenous genera, especially because such genera are mainly found in the deeper parts of the tree (e.g. Aristosyrphus, Mixogaster and Schizoceratomyia).
Despite the shortcomings, certain clades were recovered in both analyses (parsimony and Bayesian) of combined molecular and morphological characters (Table S5), most with high support, and thus form robust hypotheses. Both analyses agree also in the relatively early branching positions of Mixogaster, Schizoceratomyia, Afromicrodon and Paramicrodon.

Supporting Information
Additional Supporting Information may be found in the online version of this article under the DOI reference: 10.1111/syen.12020 Table S1. List of voucher specimens used for morphological and molecular data matrices. Table S2. List of primers used for DNA amplification. Table S3. Results of evaluations of all trees obtained with parsimony analysis of morphological characters under implied weighting for the total set of 189 taxa.