First fossil of an oestroid fly (Diptera: Calyptratae: Oestroidea) and the dating of oestroid divergences

Calyptrate flies include about 22,000 extant species currently classified into Hippoboscoidea (tsetse, louse, and bat flies), the muscoid grade (house flies and relatives) and the Oestroidea (blow flies, bot flies, flesh flies, and relatives). Calyptrates are abundant in nearly all terrestrial ecosystems, often playing key roles as decomposers, parasites, parasitoids, vectors of pathogens, and pollinators. For oestroids, the most diverse group within calyptrates, definitive fossils have been lacking. The first unambiguous fossil of Oestroidea is described based on a specimen discovered in amber from the Dominican Republic. The specimen was identified through digital dissection by CT scans, which provided morphological data for a cladistic analysis of its phylogenetic position among extant oestroids. The few known calyptrate fossils were used as calibration points for a molecular phylogeny (16S, 28S, CAD) to estimate the timing of major diversification events among the Oestroidea. Results indicate that: (a) the fossil belongs to the family Mesembrinellidae, and it is identified and described as Mesembrinella caenozoica sp. nov.; (b) the mesembrinellids form a sister clade to the Australian endemic Ulurumyia macalpinei (Ulurumyiidae) (McAlpine’s fly), which in turn is sister to all remaining oestroids; (c) the most recent common ancestor of extant Calyptratae lived just before the K–Pg boundary (ca. 70 mya); and (d) the radiation of oestroids began in the Eocene (ca. 50 mya), with the origin of the family Mesembrinellidae dated at ca. 40 mya. These results provide new insight into the timing and rate of oestroid diversification and highlight the rapid radiation of some of the most diverse and ecologically important families of flies. ZooBank accession number–urn:lsid:zoobank.org:pub:0DC5170B-1D16-407A-889E-56EED3FE3627.

Introduction larvae or adults that are currently assigned to Oestridae, Sarcophagidae, and Tachinidae [27]. However, these placements are best treated as speculative because no diagnostic characters conclusively support them [28,29,30]. It is worth noting that larvae of both Oestridae and Tachinidae are obligate endophages of living mammals and insects, respectively, with the freeliving larval stage restricted to the short period when the mature larva leaves the host to pupate. Fossilisation of such larvae, which are free-living for such a short time, must be an extremely rare event. This improbability, along with the lack of diagnostic features, suggests that pre-Quaternary fossil larvae tentatively assigned to parasitic Oestroidea are misidentified. The putative plano-convex dipteran egg attached to the pronotum of a leaf beetle (Chrysomelidae) preserved in Dominican amber [31,32] is particularly interesting. Within Diptera, the plano-convex egg shape has evolved independently in at least five lineages of arthropod parasitoids: once in the Rhinophoridae, three times in the Tachinidae (Eutherini, Exoristinae, and Phasiinae), and once in the Muscidae (Eginia Robineau-Desvoidy) [33,34,35,36]. However, while the size, shape, and position of the object on the beetle's body are suggestive of a tachinid egg, there are no other morphological clues to support or reject such a conclusion.
There are a few Quaternary (sub)fossils and copal inclusions of Oestroidea [27], but they are not discussed here because they have little bearing on oestroid evolution. Two genera in this category, Paleotachina Townsend and Electrotachina Townsend, were long thought to be Eocene fossils of Tachinidae (and were catalogued as such [27]) but were revealed as copal inclusions [37] and later assigned to the Muscidae and Sarcophagidae, respectively [30]. Not included in Table 1 are brief notes on Tachinidae and Calliphoridae in Dominican amber [31,38] or the record (with photograph) of a specimen of Calliphoridae (possibly Chrysomya Robineau-Desvoidy) from 'a piece of nearly colourless amber' [39]. The material upon which these reports were based has not been further described or illustrated and cannot be evaluated until more detailed information is published.
The absence of unambiguous pre-Quaternary oestroid fossils necessitated the dating of oestroid lineages from tectonic or other geophysical events considered causative for patterns in current distributions, like the closure of the Transantarctic corridor (minimum age for dispersal of nasal bot flies into Australia [29]), the collision of the African and Eurasian plates (maximum age for diversification of the rhino and horse stomach bot flies [29]), and the submersion of the Isthmus of Panama and the existence of a Protoantillean corridor (vicariance dating for some flesh fly lineages [40]). Nevertheless, this approach is of limited use in dating clades [41] because it assumes that the distributions of organisms at the time of these events were static and constrained by geophysical events, with little consideration for dispersal events and the widespread ebb and flow in distributions, which fossil data often indicates actually happened in most animal and plant taxa (probably in response to palaeoclimatic change) [42]. This paper describes the first unambiguous fossil of Oestroidea based on a perfectly preserved male fly discovered in amber from the Dominican Republic. We identify and characterize the specimen through digital dissection of its terminalia by CT scans, which provide morphological data for a cladistic analysis of the phylogenetic position of the fossil among extant oestroid lineages. We use the few known calyptrate fossils as calibration points for a preliminary molecular phylogeny (using 16S, 28S and CAD) to estimate the timing of major diversification events among the Oestroidea.  (Fig 1A), readily distinguishable from extant mesembrinellids by the following combination of character states: prementum about 0.65-0.70 times as long as height of head; labella broad and about 4/5 as long as prementum; palpus sub-cylindrical, about as long as antenna (Fig 1A and 1B); proepisternal depression with pale, hair-like setulae; postpronotum with three setae arranged in a triangle; scutum with 3 (presutural) + 3(postsutural) dorsocentral setae ( Fig 1C); 3 katepisternal setae; metathoracic spiracular lappet without setae; general setulae of postpronotum, scutum and scutellum relatively long and suberect; posterolateral margin of lower calypter with long trichia; stem vein bare; vein R 4+5 with fine setulae from base to about 2/3 of distance to crossvein r-m; abdominal tergite 3 with two long, erect median marginal setae and two lateral marginal setae ( Fig 1A); setae on abdominal sternites normally developed and not arranged in two rows; abdominal tergite 6 broad, not indented posteriorly and not fused to syntergosternite 7+8 (Fig 2A); cerci long, narrow and evenly curved anteroventrally (Fig 2B-2D); lateroventral lobes of distiphallus well developed (Fig 2D and 2E); scale-like vestiture of lateroventral lobes of distiphallus not visible; narrow lateral projection of lateroventral lobes of distiphallus present; acrophallus well developed, sub-cylindrical (Fig 2C and 2D).

Morphological cladistic analysis
The heuristic tree searches using parsimony with equal weights yielded 30 most parsimonious trees (MPTs) (tree length: 258 steps; Consistency Index: 0.400; Retention Index: 0.754). All of these MPTs recovered the superfamily Oestroidea (clade A) and the family Mesembrinellidae (clade C) as monophyletic, both with a Bremer support value (BS) of 4 (strict consensus tree in Fig 3). Optimization of the character transformations performed on the favoured MPT (S1 Fig)
Despite partial reliance on the notoriously inconclusive 28S ribosomal gene [17], our analyses reconstructed relationships largely consistent with those of previous studies employing considerably more data, but fewer taxa (e.g., [16,17]    3. Tachinidae: Relationships among tachinid taxa largely mirror those recovered from recent analyses using much larger data sets [17]. Tachinid monophyly is supported (41% bs,  First fossil of an oestroid fly

Morphological phylogeny of Oestroidea
Overall, the backbone relationships of oestroid families are poorly supported in our analysis. Although most families and subfamilies represented by at least two terminal taxa were reconstructed as monophyletic, the 30 most parsimonious trees vary widely in the reconstructed relationships among them (S1 Dataset). This ambiguity is apparent in the strict consensus cladogram of Fig 3,  Notably, clade P reconstructs the Tachinidae (clade Q) sister to a clade composed by Rhiniidae + (Mystacinobiidae + Rhinophoridae) (clade R; BS: 3).

Mesembrinellid diversity and systematics, and the phylogenetic position of Mesembrinella caenozoica
Mesembrinellids are a small group of 38 extant species, traditionally classified into three subfamilies and nine genera [19,43,44] (Table 2, right column). However, this morphology-based generic and suprageneric classification, which is adopted by several authors, is largely unsupported by a hierarchical array of synapomorphies [45,46,47]. Moreover, recent and ongoing studies based on molecular data [19] do not corroborate monophyly of most of the non-monotypic nominal genera of this family. Accepting genera for which explicit evidence of monophyly is lacking, inevitably leads to taxonomic and nomenclatural instability. Coupling the recent results of phylogenetic analyses of the Mesembrinellidae with such a weak generic classification would trigger the redundant reassignment of several species to different genera, or the need to erect new monotypic genera to hold 'non-fitting' species, leading to a proliferation of names and confusion. For these reasons, we have chosen to adopt a more conservative classification scheme, as recently proposed by Vargas & Wood [48] and Moll [47], by lumping all the 38 valid species of mesembrinellids under the nominal genus Mesembrinella.
In the parsimony-based phylogenetic analysis the Dominican species is reconstructed within the Mesembrinellidae as sister to Mesembrinella facialis. Clade E (M. facialis + M. caenozoica) + (M. patriciae + M. nigripes + M. perisi) includes all the representatives of the former subfamilies Souzalopesiellinae and Laneellinae sensu Guimarães [43] plus M. patriciae (Table 2), which has been recently retrieved as sister to clade M. nigripes + M. perisi also on molecular grounds [19]. Clade E is sister to the other three mesembrinellid taxa analyzed, which were formerly classified in the nominal subfamily Mesembrinellinae sensu Guimarães [43] (Table 2). These results are in contrast to the morphological phylogenies of Toma & Carvalho [45] and Bonatto [46], in which a monophyletic M. perisi + M. nigripes (as Laneella) is the basal lineage of the Mesembrinellidae, and M. facialis (as Souzalopesiella) is sister to all mesembrinellids except the clade M. perisi + M. nigripes, but the results largely agree with the recent and more comprehensive study of Moll [47]. The molecular phylogeny of Marinho et al. [19] also reconstructed M. perisi + M. nigripes + M. patriciae as monophyletic and sister to the remaining Mesembrinellidae.

Mesembrinellidae-Distribution and natural history
All extant mesembrinellid species are restricted in distribution to the rainforests of the Neotropical Region, from southern Mexico (Yucatan) to northern Argentina (Buenos Aires), with no records from the Caribbean islands except for Trinidad and Tobago [43]. The ecology of these flies is poorly known, but adults are shade-loving and occur almost exclusively within forests with a closed canopy. Occasional observations of mesembrinellids in clearings have been made during cloudy or rainy days. They are silent and fast fliers-although females fly more like big muscids, i.e., not as swift and agile as other large oestroids-mostly attracted to fermenting fruit, decaying animal matter and bird faeces [43]. Females of all species display pseudo-placental macrolarvipary, depositing one, relatively large, late first instar larva at a time, which has been nourished by secretions apparently produced by the spermathecae [49]. The few attempts of in vitro rearing of a selection of species on both animal and plant material have mostly failed (TP, unpubl.), except for Mesembrinella nigripes, which was successfully reared from first instar to adult on a variety of animal substrates, including dead snails, which were suggested as a possible primary resource for this species by Guimarães [43]. In fact, the successful breeding of M. nigripes on a diversity of decaying animal matter does not necessarily reflect its true breeding habits in the wild, but rather hints at a possible ancestral plasticity in this genus. This would be consistent with our phylogenetic reconstruction, according to which the mesembrinellids are sister to a macrolarviparous coprophage, the endemic eastern Australian Ulurumyia macalpinei Michelsen & Pape (see also the molecular phylogeny section).
According to Guimarães [43], mesembrinellids have two types of first instars that likely reflect different developmental strategies. The first instar of M. nigripes, M. facialis and M. patriciae is sub-conical or cylindrical in shape [19,43] and has slender and pointed mouthhooks, whereas first instars of other mesembrinellids (e.g., M. abaca (Hall) [43]), a shape that is unusual among oestroid flies and would seem to indicate a special diet. Character states of larval M. nigripes, M. facialis and M. patriciae are possibly plesiomorphic with respect to those of the remaining mesembrinellids [43] (but see also Toma & Carvalho [45]).

Mesembrinella caenozoica-palaeoenvironment and extinction
Fossils can reveal much about the palaeoclimatic and palaeoenvironmental conditions that prevailed at the time when the organisms lived, and can sometimes provide indirect information about the co-occurrence of species that are lacking from the fossil record. For example, the discovery of specialized fossil pollinators and phytophages indicates the presence of their pollinizers or host plants respectively, in the same way as a diversity of fossil bloodsucking arthropods suggests a diversity of vertebrate hosts, or fossils of insects with aquatic life stages suggest the proximity of freshwater [54]. Paleo-environmental reconstructions [2,31,54] based on plant and animal inclusions of the Miocene Dominican amber deposits are of a diverse tropical rain forest ecosystem with clearings, ponds and streams of the sort that occurs there today. Coupling such reconstructions with the little we know about the ecology of extant mesembrinellids would suggest that M. caenozoica was also a stenotopic, silvicolous and shade-loving species. The extinction of mesembrinellids from the Caribbean islands is noteworthy, and M. caenozoica is another instance among dozens, where a widespread clade, documented from the Caribbean through a species preserved in Dominican amber, has experienced a local extinction (or nearly so) (topic most recently reviewed in Grimaldi et al. [55]). Most such extinctions are of insect groups that presently exist in Central or South America [56], as is the case for Mesembrinellidae. One explanation for these extinctions involves insularity and the geological history of the Caribbean plate: as the Proto-Caribbean land mass drifted away from nuclear America and became insular, the biota of the various islands that today form the Greater Antilles became depauperate. However, factors other than insularity could have driven some of the extinctions through time, in particular the changing palaeoenvironmental conditions that occurred through the Late Miocene. The mesembrinellids may also have become extinct on the Caribbean islands more recently as a consequence of the Plio-Pleistocene cooling, which resulted in extensive habitat disturbance and drying [57]. The narrow ecological requirements of the mesembrinellids make them particularly sensitive to climate change, which is a major determinant of habitat loss and fragmentation.

Oestroid phylogeny: Faint light at the end of the tunnel?
Our phylogenetic analyses of Oestroidea are preliminary. The goal of our molecular analysis was not so much a focused effort to produce a robust phylogeny of the entire clade, but rather to employ available data to arrive at the best estimate of oestroid phylogeny, which served as a framework for roughly estimating ages of diversification for major lineages, based (in part) on this newly discovered oestroid fossil. In this light, we were surprised at how well these data recovered previously hypothesized clades (e.g., families, subfamilies) and how consistent many of the relationships were among different analyses (e.g., Mesembrinellidae + Ulurumyiidae; Rhinophoridae (Polleniinae + Tachinidae); Oestridae + Sarcophagidae). Still, higher 'backbone' relationships within the superfamily are not well resolved and differ among analyses, and a more intensive study employing many more loci will likely be needed to resolve them and produce a stable phylogenetic topology.
The estimated ages of key nodes in this study are largely consistent with other recent estimates of divergence times using a variety of molecular data sets, which have estimated the origin of Calyptratae at ca. 55 mya [1], 50.4 mya [11] and 60.4 mya [58]. Our estimate is similar, but somewhat older, at 67.5 my, indicating that calyptrates may date from before the K-Pg event. This similarity may stem partly from reliance on the same (few) fossil calibrations, although we are the first to include an undisputed oestroid fossil. Our results for the origin of Oestroidea (48.2 mya) and of the family Tachinidae (33.7 mya) are highly congruent with previous estimates (e.g., 56.0 and 33.2 mya [58], and ca. 40.0 and 30.0 mya [1]). A recent phylogenomic study of the insects as a whole [59] also suggests relatively recent phylogenetic origins for calyptrates and oestroids in the Upper Paleogene and Lower Oligocene, respectively. Remarkably, however, this study suggests that the split between Sarcophagidae and Tachinidae is approximately at the same depth as that between the apine genera Apis Linnaeus and Bombus Latreille (Hymenoptera: Apidae), if not younger. Regardless of the exact time of phylogenetic origin, the Calyptratae have likely experienced multiple episodes of rapid radiation, diversifying into about 22,000 described (and many undescribed) species in a relatively short time [1]. It may not be a coincidence that the origin of the calyptrates is estimated to be around the Cretaceous-Palaeogene mass extinction event. The upheaval of existing ecological communities and the opening of new niches following such a massive extinction event may have facilitated the diversification of lineages and life histories in this ecologically plastic clade of flies, as has been suggested for mammals, passerine birds, and other taxa [60,61].
The mesembrinellid clade (including Ulurumyia macalpinei) appears to be of late Eocene age (43 myo), with the Mesembrinellidae estimated as about 39 myo, substantially older than the fossil of M. caenozoica. Given the uncertainty about where the fossil taxon fits among other Mesembrinella species in our molecular phylogeny, it is difficult to ascertain where the fossil calibration constraint should be placed on the phylogeny. It is possible that constraining a subgroup to a minimum age of 15 my, rather than the whole family, could push our estimated node ages further back. The near-basal position of Mesembrinellidae within the Oestroidea, isolated from other lineages of Calliphoridae sensu lato, was suggested previously by Crosskey [50] and Pape [62], although the close association with Ulurumyiidae is a novel hypothesis.
Morphology and molecules have usually yielded markedly different phylogenetic reconstructions of the oestroid lineages, and this applies to our analyses as well. Two exemplar cases involve the sister group relationship of the Tachinidae and the monophyly/non-monophyly of the bot flies. All recent phylogenetic reconstructions based on multiple genes [16,17,63] converge in reconstructing the calliphorid genus Pollenia as sister to a monophyletic Tachinidae with strong statistical support. Remarkably, there are very few clues supporting this hypothesis on morphological or ecological grounds, except that larvae of both Pollenia and 'lower' tachinids [36,64] appear to be parasitoids of soil-dwelling organisms (i.e., earthworms and larvae of litter-associated weevils, respectively) [64]. On the other hand, molecular-based reconstructions often fail in retrieving Oestridae as monophyletic ( [16,17] in part), ( [18] in part), [19], despite strong support from morphological data [12,13,28,29] (Fig 3). As remarkable exceptions, all analyses presented here agree in reconstructing the family Oestridae as monophyletic and, for the first time, Ulurumyia macalpinei as sister to Mesembrinellidae, with this clade in turn being sister to the remaining Oestroidea (or at least much of it, Figs 3 and 5).
This situation exemplifies the uncertainty and difficulty in reconstructing relationships among oestroid lineages. Disagreement among reconstructions could be attributed to a possible rapid diversification event of persistent lineages [1,17]. However, our time-calibrated phylogeny suggests that oestroid diversification as a whole may have been relatively 'slow' but punctuated by multiple episodes of more rapid radiation throughout the Cenozoic (Fig 5). On the other hand, the early evolution of oestroids could have been a time of 'evolutionary experimentation' during which many early lineages went extinct, leaving a somewhat skeletal phylogeny with only a few surviving lineages that later diversified. Without a better fossil record, it is impossible to discern between these possible scenarios.
Phylogenetic trees inferred from multiple gene sequences are starting to converge into similar, more stable patterns [59,65,66,67,68], and this is true also for the oestroids despite the sparse and uneven taxon coverage [16,17]. If we assume that the recent molecular data are better at recovering the true phylogeny of Oestroidea, then we are still left with the question of why the morphology is so misleading and what this suggests about character evolution in this group. Conversely, if we assume inferences from the morphological data are closer to the truth, why are molecular-based inferences less accurate? Presently, we have a limited set of morphological characters with which to infer the phylogeny of the Oestroidea and this is due to the slow pace of detailed morphological investigation in recent decades. For many clades there are only one or two supporting character states, which are often homoplastic (S1 Fig). From this, it follows that very few additional characters, alternative homology assessments, or the use of different methods, might cause substantial topological changes to the inferred phylogenies. Also, there is no simple way to assess the complexity of morphological character evolution, except by their implied weights based on the number of homoplastic occurrences (but see [69]).
Differences between molecular-and morphology-based reconstructions of Oestroidea may seem unbridgeable, but the discovery of the first, undisputed oestroid fossil allows us to better establish the age and the 'tempo and mode' of diversification in these flies. Our results suggest that the K-Pg extinction event may have played a crucial role in boosting calyptrate diversification through the Cenozoic, as it did for the major radiations of birds, mammals and angiosperms.

Amber inclusion
The amber inclusion was acquired from Mr. Jorge Martínez of Santiago, Dominican Republic, whose well-known workshop and business has provided Dominican amber inclusions for scientific study for decades. The mine source of the specimen is impossible to determine with certainty, since rough amber from various mines located north-northeast of Santiago [70] is usually mixed during processing in the workshops in Santiago. It is certain that the specimen is not copal (subfossilized resin that is only hundreds to thousands of years old, outcrops of which occur in eastern Dominican Republic in the vicinity of Cotui) because it is not reactive with organic solvents and is not the typical very light yellow colour of copal. The inclusion with the male holotype of Mesembrinella caenozoica sp. nov. is in the James Zigras collection, housed in the Division of Invertebrate Zoology at the American Museum of Natural History, New York.

CT-scan analysis
Scanning was done in the Microscopy and Imaging Facility (MIF) at the American Museum of Natural History, using a General Electric Phoenix v|tome|x-s nanotube high resolution computed tomography system with a tungsten target, an accelerating voltage of 90kV and a current of 200μA, at resolutions between 9 and 7.5 μm. 300ms exposures were taken at 0.2˚intervals. Volumetric data was produced with GE Phoenix's Datos Reconstruction 2.2.1 software using a cone-beam filtered back projection reconstruction algorithm. A beam hardening correction was applied. Volumetric data was exported as16-bit integer greyscale TIFF stacks. Image stacks were combined in the Fiji distribution of NIH's ImageJ with the Pairwise 3D stitching plug-in. Imaging focused on the male terminalia, in order to visualize hidden but informative features of the phallus such as the dorsolateral processes, and the cerci. Data were rendered and animated using Volume Graphics Studio Max 2.2.6 software equipped with the Coordinate Measurement Module.

Morphological cladistic analysis
To evaluate the phylogenetic position of Mesembrinella caenozoica sp. nov. among extant Oestroidea a matrix was constructed of 74 morphological characters for 49 taxa (S1 Dataset), First fossil of an oestroid fly mostly adapted from previous studies [12,13,14] with minor adjustments from our own studies. Oestroid diversity is represented by a sample of 45 species, including M. caenozoica described here, plus four muscoid outgroups (three Muscidae, one Anthomyiidae). The sample includes all oestroid subfamilies, except Paramacronychiinae (Sarcophagidae) (see S2 Text for a complete list of included taxa and depositories of material studied).
The data matrix was produced in Mesquite version 3.03 [72] (S1 Dataset). Cladistic analysis was conducted with TNT version 1.5 [73,74]. Heuristic searches were run with the 'traditional search' option with the following parameters: General RAM of 1 GB, memory set to hold 1,000,000 trees, setting 1,000 replicates with tree bisection-reconnection (TBR) branch swapping and saving 1,000 trees per replicate. Multistate characters were treated as unordered and zero-length branches were collapsed. Inapplicable and unknown states were coded as '-' and '?', respectively, in Mesquite. Character state changes were optimized in WinClada version 1.00.08 [75]. The favoured tree was selected among the most parsimonious trees by calculating the total fit (command 'fit Ã ' in TNT) for every tree under a range of k-values (S3 Text), using the unambiguous transformation algorithm, and choosing one of the trees with highest fit. Bremer support values were calculated in TNT from 15,000 trees up to 10 steps longer than the shortest trees obtained from a 'traditional search', using the 'trees from RAM' setting. Consistency and retention indices were calculated in TNT.

Molecular phylogenetic analyses and divergence time estimation
Sequence data for 89 taxa was obtained either directly via PCR amplification of extracted DNA from collected specimens or through deposited sequence data from NCBI Genbank (see S2 Text). Most 28S and 16S sequences of mesembrinellids and some other oestroids were from Marinho et al. [15,19], and most tachinid sequences, as well as a number of CAD sequences of other taxa, were from Winkler et al. [17]. For the newly obtained sequences, flies were collected by hand netting and 1-3 legs were removed from each specimen and placed in 95% ethanol shortly after collection. DNA extraction and isolation were performed with a Puregene1 Tissue Kit (Qiagen Inc.) using standard methods (see [17]). PCR amplifications employed the primers 28SF and 28SR for the 5' half of the 28S rDNA gene and 54F and 405R for the 5' region of the CAD gene ('CAD1'; see Winkler et al., [17], Moulton & Wiegmann [76]). PCR amplification protocols and conditions followed those outlined in Winkler et al. [17].
Sequences from each locus were aligned separately with MAFFT v.7 [77] via the online MAFFT server (http://mafft.cbrc.jp/alignment/server/), employing the G-INS-1 algorithm for CAD and the L-INS-I algorithm for 28S and 16S. A large unalignable intron of variable length in the centre of the fragment was trimmed from CAD [76]. Final sequence lengths were 1650 bp, 1615 bp, and 665 bp for 28S, CAD, and 16S respectively. In a few cases, concatenated sequences of different genes were from different congeneric species (see S2 Text). The full alignment is available through the online data repository TreeBASE (treebase.org).
A Maximum Likelihood analysis was performed using GARLI v2.01 [78]. The concatenated data set was partitioned by locus, and within CAD by codon position, and a separate GTR+I +Γ substitution model was defined for each partition, following the best partition and substitution model scheme given by PartitionFinder v1.1.1 [79], with all parameters estimated from the data. GARLI run was set as follows: 5 independent search replicates, 5x10 6 generations and default options for automated stop. The hippoboscoid Glossina morsitans Westwood was selected as outgroup. 100 non-parametric bootstraps of the data set were conducted using the original search parameters, except for the genthreshfortopoterm option, which was set to 10,000. Similar results were obtained in a RaXML-HPC v. 8 [83], both also conducted on XSEDE servers via the CIPRES web portal (v. 3.3) [81]. MrBayes analysis was conducted with the same partition and substitution model scheme used in the ML analysis and the run was set as follows: 50x10 6 generations with two sets of 6 chains, sample frequency = 1,000 and burn-in set to 25% after checking for convergence. Node supports were assessed by analysing the posterior probabilities in the 50% extended majorityrule consensus tree. The BEAST analysis was set similarly to the MrBayes run and additional priors (for divergence time estimation) included a Yule coalescent model with parameters estimated with a gamma prior for relative birth rate, gamma shape parameters with exponential priors, and all rates with gamma distributed priors using default values of alpha and beta. The hippoboscoids (Glossina morsitans and Ornithoctona erythrocephala (Leach)) were assigned as the outgroup (i.e., all other taxa were constrained to be monophyletic). A relaxed log-normal clock model was used to estimate divergence times [84,85], with an exponential distributed mean prior (ucldMean.c; mean = 10) and a gamma distributed standard deviation. Four calibrations were used at four nodes in the tree, in each case using a log normal distributed prior for the estimated tmrca (time to most recent common ancestor), and all were assumed to define a monophyletic group. These calibrations included: 1. Phaoniinae (Muscidae) from 15-20 myo Dominican amber [22], used to constrain Phaonia Robineau-Desvoidy + Cordiluroides Albuquerque (initial prior offset = 15); 2. Glossina Wiedemann (Glossinidae) from 35 myo Colorado shale [2], used to constrain Glossina + Ornithoctona Speiser (offset = 35); 3. Protanthomyia Michelsen (stem group Anthomyiidae) [20] from ca. 42 myo Baltic amber, used to constrain Anthomyiidae (offset = 42), and the current fossil Mesembrinella from 15-20 myo Dominican amber used to constrain the minimum age of Mesembrinellidae as a whole (offset = 15). Prior parameters for each of these distributions were estimated with log-normal hyper-prior distribution, beginning with M = 2.0, and S = 2.50. MCMC chain length was 50x10 6 generations with trees stored every 1,000 generations and parameters logged every 1,000 generations. A conservative burn-in frequency of 25% was used based on visual examination of tree likelihood convergence using TRACER. TreeAnnotator v2.4.0 was used to calculate the maximum clade credibility tree and posterior probabilities of nodes based on the 37 500 trees retained by the MCMC analysis.

Nomenclatural Acts
The electronic edition of this article conforms to the requirements of the amended International Code of Zoological Nomenclature, and hence the new names contained herein are available under that Code from the electronic edition of this article. This published work and the nomenclatural acts it contains have been registered in ZooBank, the online registration system for the ICZN. The ZooBank LSIDs (Life Science Identifiers) can be resolved and the associated information viewed through any standard web browser by appending the LSID to the prefix "http://zoobank.org/". The LSID for this publication is: urn:lsid:zoobank.org:pub: D808DBAD-EE10-466F-94E5-A4B536EE938A. The electronic edition of this work was published in a journal with an ISSN, and has been archived and is available from the following digital repositories: PubMed Central, LOCKSS.