A multilocus phylogenetic framework of the tribe Aeromachini (Lepidoptera: Hesperiidae: Hesperiinae), with implications for taxonomy and biogeography

The Oriental tribe Aeromachini (Lepidoptera: Hesperiidae: Hesperiinae) is diverse and widespread, but its monophyly and circumscription remain unresolved. In this study, we inferred phylogenetic relationships within the tribe using two mitochondrial (16S and COI‐COII) and two nuclear genes (EF‐1α and Wingless) from 71 samples representing all of the known genera of Aeromachini (s.l.). Phylogenetic topologies obtained from two model‐based methods (maximum‐likelihood and Bayesian inference) were largely congruent, showing that Aeromachini was monophyletic. Many genera within Aeromachini were polyphyletic, especially Arnetta, Thoressa, Ampittia and Sovia. Based on these results, we propose a new classification consisting of 14 genera, of which two are newly described: Halpemorpha gen.n. and Praethoressa gen.n. Ampittia and Ochus are restructured, with: (i) Ampittia nana and Ampittia dalailama moved to Aeromachus, (ii) new combinations proposed for the remaining Ampittia species, and (iii) a new combination, Ampittia subvittatus comb.n., proposed. The Oriental members of the genus Arnetta belong to this tribe, but the African Arnetta hyposticta was found to be embedded into the outgroup and closely related to Isoteinon. We propose that hyposticta be reinstated in the genus Galerga Mabille, along with two other African species (fito and ellipsis). Thoressa is polyphyletic, with three distinct lineages. The type species Thoressa masoni forms a clade closely related to the sister genera Sebastonyma and Parasovia. We thus describe Praethoressa gen.n. to include the species varia (formerly of Thoressa), and transfer the remaining Thoressa to Pedesta, which is confirmed as a valid genus. Although the relationships of some genera remained unsolved in our work, the present phylogenetic hypothesis will serve as a reference for further studies. A secondary calibration was used to estimate the divergence time, which indicated that the diversification of Aeromachini started in the late Eocene and early Oligocene (c. 34.22 Ma). Further, ancestral range estimates indicated that the common ancestor of Aeromachini originated in Southeast Asia. Diversification in situ and range expansion to adjacent areas have played essential roles in the evolution of Aeromachini, but founder‐event speciation and vicariance also have likely been important factors in shaping the biogeographical history of this tribe.


Introduction
Considering the limits to morphological classification methods, robust molecular phylogenies can contribute to the development of more stable classifications and provide better frameworks for recovering the patterns of diversification and historical biogeography. The family Hesperiidae is well-represented in all zoogeographical regions, but the phylogeny of this family and other biological aspects, such as utilization of host plants, biogeography and adaptive evolution, are much better understood for the European and American lineages than for the existing Asian and African lineages. The classification of Hesperiidae has been revised based on molecular analyses alone or in combination with morphological characters, but its higher-level classification is still uncertain (Warren et al., 2008(Warren et al., , 2009Sahoo et al., 2016). The largest subfamily Hesperiinae has been shown to be monophyletic, but the tribe-level systematic associations of many genera within this subfamily remain ambiguous. The tribe Aeromachini (Hesperiidae: Hesperiinae) is currently treated as a monophyletic group within Hesperiinae. However, tribal delimitation and generic inter-relationships are inconclusive and inadequately demonstrated at present. The tribe Aeromachini (Lepidoptera: Hesperiidae: Hesperiinae) is one of the most diverse lineages of Hesperiinae. The tribe currently comprises approximately 130 described species in 12 genera, which are predominantly distributed in the Oriental Region, with some species in the Palaearctic Region and some Ampittia species distributed in Madagascar (Evans, 1937(Evans, , 1949Devyatkin, 1996;Warren et al., 2009;Cock & Congdon, 2012;Yuan et al., 2015a). Members of this tribe are small-to medium-sized, usually with yellowish to dark brown wing colour, and with a brand or stigma on the forewing of males. Vein CuA 2 of the forewing arises opposite to or beyond the origin of vein R, closer to CuA 1 than to M 3 . The antennae are less than half the length of the forewing costa (Evans, 1949;Eliot, 1992;Chou, 1998;Warren et al., 2009, Yuan et al., 2015a. Larval hostplants are generally from the grass family Poaceae (Inoué & Kawazoé, 1966;Hsu, 1999Hsu, , 2015Igarashi & Fukuda, 2000;Lyu & Chen, 2015), except that some species feed on Malvaceae (Saji & Valappil, 2017).
The tribe Aeromachini s.l. has been a taxonomically unresolved group due to the disputed generic and specific placements in this taxon and their uncertain taxonomic validity. It also has been called the Astictopterus group (Evans, 1949), Halpe group (Inoué & Kawazoé, 1966), Ampittiini and Astictopterini (Chou, 1994(Chou, , 1998 or Halpini (Yuan et al., 2015a). Tutt (1906) originally introduced Aeromachinae as a subfamily and Aeromachidi as a tribe for genera Taractrocera Butler, Ampittia Moore and Aeromachus Nicéville, although Taractrocera is currently placed in the tribe Taractrocerini. According to articles 23.3.1 and 34.1 of the Code of the International Commission on Zoological Nomenclature (ICZN, 1999), the designation Aeromachini is the valid one. Despite the variety of names and taxonomic designations applied to this taxon, the disagreements among authors are concerned mainly with the systematic positions of Astictopterus and Arnetta. The genus Astictopterus was successively placed into the Astictopterus group (Evans, 1949), Isoteinon group (Inoué & Kawazoé, 1966), Ancistroides group (Eliot, 1992), Astictopterini (Chou, 1994(Chou, , 1998 and Ancistroidini (Yuan et al., 2015a). Moreover, Warren et al. (2008) showed that Astictopterus was closely related to Kedestes, and these two genera were closely associated with Isoteinon, suggesting that Astictopterus does not belong to Aeromachini. This result was reinforced by combined molecular and morphological data (Warren et al., 2009). Evans (1949) treated the other poorly resolved genus Arnetta as a member of the Astictopterus group based mainly on the following morphological characters: veins Sc and R apart, apiculus obtuse or hooked, and the upperside of forewing with or without hyaline spots. He pointed out that the tuft on the dorsum and an obscure brand were present in males of only two species, Arnetta atkinsoni and Arnetta verones. Subsequently, Inoué & Kawazoé (1966) placed Arnetta in their Halpe group based on the porrect palpi of three Malaysian specimens of Arnetta atkinsoni. Eliot (1978Eliot ( , 1992 noted that Arnetta was probably not monophyletic and provisionally transferred it to the Plastingia group based primarily on the shape of wings and the slightly oblique direction of vein M 2 on the hindwing. Warren et al. (2009) agreed that Arnetta should not be included within Aeromachini and listed it as incertae sedis due to the absence of the 'Halpe brand'. However, there is no molecular evidence in support of most of these arrangements. Yuan et al. (2015a) noted that Arnetta resembled Iambrix and Idmon in some characters, but treated it as a member of Isoteinonini. Thus, the circumscription of the tribe Aeromachini is uncertain.
Although Warren et al. (2008) presented a revised classification of the family Hesperiidae based on large-scale sampling of molecular data, the infrafamily phylogenetic relationships are still unresolved, particularly for the Old World taxa. Warren et al. (2008Warren et al. ( , 2009) included only four out of the 12 known genera of Aeromachini in their analysis, and concluded with strong phylogenetic support that the tribe is monophyletic and basal within Hesperiinae. A recent study (Sahoo et al., 2016) built on the findings of Warren et al. (2009) and corroborated the monophyly of Aeromahini, although without examining additional specimens. Yuan et al. (2015b) divided the Chinese Hesperiidae into five subfamilies, but did not recover Aeromachini as a monophyletic group, perhaps because Halpe, Pithauria and Aeromachus were represented by only one species each. Thus, although some attempts have been made to improve the situation (Dodo et al., 2008), phylogenetic relationships within the tribe Aeromachini are still poorly resolved.
Due to the scarce hesperiid fossil record (de Jong, 2016), little of the historical biogeography of this tribe is understood. Sahoo et al. (2017) reconstructed a time-calibrated phylogeny to explore what factors affected the diversification of skipper butterflies (Hesperiidae) using previously estimated ages as a split range, and a fossil as a minimum stem age. Most recently, a more comprehensive time-calibrated phylogeny of butterflies (Papilionoidea) was proposed, which used 12 fossils and 10 host plant nodes for calibration (Chazot et al., 2018). Such frameworks provide an opportunity to illuminate the history of different butterfly groups. The development of a similar framework for the Aeromachini provides useful information for the study of this group.
The present study aimed to: (i) propose an alternative hypothesis of generic relationships within the Aeromachini with additional taxon sampling and multi-locus mito-nuclear phylogenetic analyses; (ii) clarify taxonomic status of some genera listed as incertae sedis by Warren et al. (2008Warren et al. ( , 2009; and (iii) provide a preliminary hypothesis of the historical biogeography of this tribe.

Taxonomic sampling
We sampled all of the genera listed in the Astictopterus Group (Evans, 1949), and Parasovia as described later. We included the type species for each genus except for Baracus vittatus, Onryza meiktia and Pithauria murdava, in addition to as many species as possible from each genus. We sampled a total of 71 individuals belonging to 46 species in 13 genera as ingroup taxa, and 10 species from other subfamilies and related genera within the Hesperiinae as outgroup taxa. We sampled and sequenced most of the individuals used in this study except for 24 sequences from eight samples obtained from Genbank. Table S1 gives a checklist of specimens with the collecting localities and voucher codes. We deposited the voucher specimens in the Insect Collection of the South China Agricultural University (SCAU), except for the Indian samples, which were deposited in the research collections of the National Centre for Biological Sciences (Bangalore, India).

DNA extraction, PCR amplification and sequencing
We extracted genomic DNA using the Qiagen DNeasy Blood and Tissue Kit (Qiagen, Germany) or HiPure Insect DNA Kit (Magen, China) from two or three legs of dried adult specimens or specimens preserved in pure molecular-grade ethanol following slightly modified protocols recommended by the manufacturers.
We sequenced four loci commonly used in butterfly phylogenies: mitochondrial Cytochrome c oxidase subunit I and II (COI-COII, 931 bp) and 16S ribosomal RNA (16S rRNA, 591 bp), and nuclear Elongation factor 1 alpha (EF-1 , 1066 bp) and Wingless (400 bp). We excluded 89 bp of 16S rRNA from subsequent analyses due to ambiguous regions (see below). Table 1 lists the PCR primers.
We performed most PCR amplifications in 20 L reactions containing 10 L KOD PCR Buffer (Toyobo, Osaka, Japan), 4 L 2× dNTPs, KOD FX, 0.5-1 L of each primer and 1 L of the DNA extract. The amplification cycle for COI-COII and 16S was 2 min at 94 ∘ C, and 35 or 40 cycles of 10 s at 98 ∘ C, 0.5 or 1 min at 47 ∘ C (COI-COII), or 48 ∘ C (16S) and 1.5 min at 68 ∘ C. For EF-1 and Wingless, the initial 94 ∘ C at 2 min was followed by 35 or 40 cycles of 10 s at 98 ∘ C, 55 ∘ C (EF-1 ) or 58 ∘ C (Wingless) for 1 min and then 1.5 min at 68 ∘ C. All amplification cycles included a final extension period of 68 ∘ C for 10 min. We visualized the results by electrophoresis in 1% agarose gel stained with ethidium bromide. We sequenced the PCR products after cloning into pMD 18-T vector (TAKARA) referring to Fan et al. (2016). We compared all of our sequences using blastn in GenBank to check for cross-contamination. Our new sequences are deposited in GenBank (Table S1).

Phylogenetic inference and divergence time estimate
We aligned all sequences, except the 16S ribosomal RNA sequences, using Clustal W (Thompson et al., 1997) implemented in MEGA 7.0 (Kumar et al., 2016), with default parameters. For 16S rRNA, which contained ambiguous regions of alignment representing hypervariable regions (HVRs) unlikely to evolve on a per-site nucleotide substitution basis, we aligned sequences online with MAFFT v7 (http://mafft.cbrc .jp/alignment/software/; Katoh & Standley, 2013) by applying a secondary structure model with a stem candidate length of 2 and a threshold of base pairing probability of 0.01. We identified and excluded the ambiguously aligned regions using Gblock v0.91b (Castresana, 2000;Talavera & Castresana, 2007), allowing for smaller final blocks, gap positions within the final blocks and less strict flanking positions. We treated gaps in all analyses as missing data.
We used PartitionFinder v2.1.1 (Lanfear et al., 2016) to select the optimal partitioning scheme under the Bayesian Information Criterion (BIC). We performed phylogenetic analyses using maximum-likelihood (ML) and Bayesian Inference (BI). We performed ML analyses using IQ-TREE 1.5 (Nguyen et al., 2015) as implemented on the W-IQ-TREE web server (http://iqtree.cibiv.univie.ac.at/; Trifinopoulos et al., 2016). We used the partition scheme produced by PartitionFinder, but the best fitting model was selected using ModelFinder (Kalyaanamoorthy et al., 2017) implemented in IQ-Tree because it has more sequence evolution models than PartitionFinder. We estimated models of substitution using Auto function with FreeRate heterogeneity. We evaluated the node support by 1000 ultrafast bootstrap replicates (UFBS) (Minh et al., 2013). We also conducted standard ML analyses using RaxML (Stamatakis et al., 2008) on CIPRES Science Gateway (Miller et al., 2010) under the GTR+GAMMA model, using the partition scheme from the PartitionFinder analysis. We performed bootstrapping with 1000 replicates under auto Majority Rule Criterion (autoMRE).
We used BEAST v2.4.8 (Bouckaert et al., 2014) for BI to simultaneously investigate phylogenetic relationships and estimate the divergence times. Calibration data were taken from the divergence time estimate of butterflies (Papilionoidea) (Chazot et al., 2018). The second calibration point was set as an age 28.67 Ma (95% highest posterior density, , to the timing of the split between Halpemorpha and Pedesta. We employed the partition scheme and models produced by PartionFinder. We applied the uncorrelated lognormal relaxed clock model and the Yule model as tree priors. The prior for the hyperparameter 'ucldMean' was assigned as an exponential distribution of mean 10.0. We specified an Markov Chain Monte Carlo (MCMC) chain length of 100 million. We estimated clade robustness by posterior probabilities (PP) of BI and ultrafast bootstrap values (UFBS, IQ-Tree) in ML. We deemed branch support to be strong when PP was 0.95-1 and UFBS was 95 or higher, moderate when PP was 0.90-0.95 and UFBS 90-95, and weak/low when PP was 0.50-0.90 and UFBS less than 90.

Ancestral range estimation
We pruned the MCC tree using the R (R Core Team, 2013) package APE (Paradis et al., 2004) to obtain a phylogeny with only one tip per species of Aeromachini spp. We recognized and coded six distribution areas based on biogeographical realm and geotectonic plates, with some modifications according to the certain distributions of species, as follows: ( (Hall, 2012(Hall, , 2013 and that there were few endemic species in each area, the time slices methods were not applied in these analyses. The maximum number of ancestral areas was set to three, as no extant species occurs in more than three biogeographical regions. Ancestral areas inferences without constraints were performed with the R package BioGeoBEARS (Matzke, 2014) for the tribe Aeromachini. We tested each of the implemented models DEC, DIVALIKE and BAYAREALIKE, both with and without the additionally implementation of J (Matzke, 2013), which specifically donated the parameter representing the process of 'jump dispersal' or 'long-distance dispersal'. The log-likelihood (LnL) and Akaike Information Criterion (AIC) were used to select and evaluate the best model.

Dataset properties
The final dataset for the molecular phylogenetic analysis consisted of 81 taxa representing 56 species, of which 10 were outgroups, and representing all of the genera within the tribe Aeromachini. The resultant alignment of the four mito-nuclear genes consisted of 2899 bp, of which 1213 were variable and 917 parsimony-informative sites.

Phylogenetic inference
File S1 lists the best partitioning schemes and the optimal evolutionary models from PartitionFinder for each of the partitioned datasets. File S2 lists the specific models for each partition matrix from IQ-Tree. IQ-Tree and RaxML runs produced similar topologies ( Fig. 1 and File S3). We present below the results from IQ-Tree. The BI tree had congruent topologies to ML trees within Aeromachini, but some nodes were different among the outgroup taxa (File S4). Nodes supported strongly in ML analyses were always recovered with strong posterior probabilities, expect that the basal support was low (File S4). Both BI and ML analyses supported the monophyly of Aeromachini sensu Warren et al. (2008) (Fig. 1 and File S4). Within the tribe, Baracus, Halpe and Pithauria were each found to be monophyletic with strong support, whereas the remaining genera were either paraphyletic or polyphyletic.
The genus Arnetta was polyphyletic, which divided into two distantly related clades. The clade including two Asian species, Arnetta atkinsoni (the type species) and Arnetta verones, was sister to the rest of the Aeromachini. Arnetta hyposticta from Madagascar showed a close relationship with Isoteinon, and the relation was strongly supported (UFBS = 97, PP = 0.99).
Sovia also was polyphyletic and divided into two unrelated clades: the Sovia clade and Clade A. The Sovia clade, including S. lucasii and S. subflava, was strongly supported (UFBS = 100, PP = 1). Clade A also was strongly supported as a monophyletic group (UFBS = 100, PP = 1), consisting of S. albipectus and S. eminens, and sister to the remaining taxa except the genera Arnetta, Sovia and Aeromachus. Accordingly, we describe this clade (Clade A) below as a new genus, Halpemorpha gen.n.
The genus Ampittia was not recovered as a monophyletic group. The clade including the type species Ampittia dioscorides clustered together with the monotypic genus Ochus with strong support (UFBS = 100, PP = 1), but the inter-relationships within this clade remained uncertain. The other subclade comprising two species, Ampittia nana and A. dalailama, clustered together with Aeromachus with strong support (UFBS = 100, PP = 1).
Although the monophyly of Halpe was strongly supported, the inter-relationships remained unresolved. Its three species (H. porus, H. nephele and H. gamma) formed a well-supported subclade. Halpe pelethronix and H. lutisquama were recovered as sister species in the BI tree, but this topology was not obtained in ML analyses. Species of Thoressa formed four distinct clades. The type species, Thoressa masoni, was the sister taxon to a clade composed of Parasovia and Sebastonyma, with strong support (UFBS = 99, PP = 1). Another representative of this genus, Thoressa varia from Japan, was placed as a sister group to Pithaura with strong support (UFBS = 100, PP = 1). We recognize and describe this clade as a new genus, Praethoressa gen.n. The remaining Thoressa species clustered together with species of Pedesta with moderate support in ML and strong support in BI (UFBS = 90, PP = 0.98). Two samples of Thoressa hyrie from India formed a weakly supported isolated subclade that was a sister clade to another subclade that included Pedesta masuriensis, P. blanchardii and P. xiaoqingae. Pedesta baileyi was nested within the Clade B that included five Thoressa species: T. latris, T. fusca, T. bivitta, T. gupta and T. submacula.

Divergence time and ancestral range estimation
We used Tracer v1.6 to check the BEAST log file, which showed that all parameters had ESS >300, and most were >1000. The MCC trees resulting from BEAST analyses generated congruent topologies with our phylogenetic analyses (as aforementioned). The divergence time estimates indicated that the diversification of the Aeromachini started in the late Eocene and early Oligocene ( Fig. 2; (Table 2) based on the one tip per species chronogram (excluding outgroup). We thus present and discuss the results from this analysis herein. The common ancestor of the Aeromachini was estimated to have originated in Southeast Asia (Fig. 3). Four founder-event speciation events were estimated during the mid-Miocene: A. plumbeola, and H. luteisquama colonized Philippines via long-distance dispersal at c. 11.14 and 11.56 Ma, respectively; P. varia colonized Japan and Sakhalin (JS) at c. 11.57 Ma; the genus Baracus was inferred to originate in the Indian Subcontinent and diversified there in the late Miocene (c. 7.74 Ma); and the genus Pedesta split into two lineages at c. 11.44 Ma -one of colonized and diversified in SWC at c. 9.95 Ma and the other diversified in situ and dispersed to adjacent areas. The most probable ancestral species of all other genera within Aeromachini likely occupied Southeast Asia and diversified in adjacent regions around the mid-Miocene.

Systematics of Aeromachini
The dense and extensive taxon sampling that we used and the resultant molecular evidence confirmed that the tribe Aeromachini (sensu Warren et al., 2008) is a monophyletic group. Relative to the previous analyses (Warren et al., 2008(Warren et al., , 2009Yuan et al., 2015b), the present study has resulted in considerable resolution at the genus level within this tribe. However, the statistical support for the stem of the tribe was not strong in our phylogenetic analyses, and the inter-relationships and monophyly of some genera still remained poorly resolved across all the analyses, and thus requires further investigation to resolve. On the basis of the phylogenetic results and morphological characteristics, we recognize 14 genera within Aeromachini, of which seven are largely consistent with traditionally recognized genera as already described or with little revision (Baracus, Halpe, Pithauria, Onryza, Sebastonyma, Parasovia, Aeromachus), two are proposed as new (Halpemorpha gen.n. and Praethoressa gen.n.) and the remaining five are inconsistent with previous generic concepts (Arnetta, Ampittia, Sovia, Pedesta and Thoressa). All of the taxonomic changes proposed in the present study are listed in File S5. Below we provide detailed taxonomic notes on the significant genera within the tribe Aeromachini, and describe in detail the two new genera.

Genus Arnetta Watson, 1893
The genus Arnetta, a small genus with seven species (three from Madagascar and four distributed from India to Borneo), has an ambiguous taxonomic status due to the paucity of genus-specific diagnostic characters. Evans (1937Evans ( , 1949 defined the genus mainly based on the wing pattern and ignored the extensive differentiation in the morphology of the male genitalia among species. Due to its unclear morphological definition, subsequent researchers continued to change the taxonomic position of Arnetta based on some particular traits, which, according to our results, are not apomorphic characters of this genus. The African species hyposticta is not congeneric with two Asian species, atkinsoni and verones, especially with respect to their  male genitalia. Uncus is not separated and gnathos is stout in hyposticta, whereas uncus is deeply bifurcated at the tip and gnathos is slender and almost reaches the tip of uncus in the latter two species (Evans, 1937(Evans, , 1949. Eliot (1978Eliot ( , 1992 realized that the genus was probably not a monophyletic group, but the lack of sufficient specimens and species for examination resulted in placing the genus in the Plastingia group. Our phylogenetic analyses indicated that these two Asian species of Arnetta belong to the Aeromachini, which is unlike the classification in previous studies (Warren et al., 2009;Yuan et al., 2015b), but partially agrees with Inoué & Kawazoé (1966). Another species in our phylogeny, A. hyposticta (the type species of Galerga), shows a close relationship with the genus Isoteinon, but there are strong morphological dissimilarities between the genera Galerga and Isoteinon. Galerga (type species Galerga hyposticta Mabille) was treated as a synonym of the genus Arnetta by Evans (1937). Lees et al. (2003) treated Galerga as a subgenus of Arnetta, and included two species, Arnetta (Galerga) hyposticta and A. (G.) ellipsis. Following Viette's (1956) suggestion, we here reinstate the genus Galerga as valid, and reinstate Galerga hyposticata Mabille, rest.stat. Two other African species, ellipsis Saalmuller and fito Evans, are similar to hyposticta in term of the wing pattern and genitalia (Evans, 1937;Viette, 1956). Hence, we reinstate ellipsis and fito as valid species under Galerga, as Galerga ellipsis (Saalmüller), rest.stat. and Galerga fito (Evans, 1937), rest.stat. As noted by Evans (1949), two other Indian species (Arnetta mercara Evans and Arnetta vindhiana Moore) resemble Astictopterus jama in external features (also see Kunte et al., 2017) but their male genitalia are closer to those of Galerga than to those of Arnetta and Astictopterus jama. Thus, it seems reasonable to place them in Galerga, although molecular phylogenetic evidence will be useful to confirm this hypothesis.
In summary, we retain only the species verones and atkinsoni in the genus Arnetta. The other five species previously assigned to this genus should be members of Galerga. The systematic positions of the species mercara and vindhiana need further research. Based on the molecular evidence, as well as morphological characters, we conclude that Arnetta is a subordinate taxon within Aeromachini.
Genus Sovia Evans, 1949Evans (1949 recognized six species and several subspecies under Sovia. Subsequently, four new species were described and a subspecies was elevated to species level (Devyatkin, 1996;Huang, 2003;Huang & Wu, 2003;Miyazaki & Saito, 2010;Xue et al., 2015). To date, Sovia consists of 11 species that are exclusively distributed in South and Southeast Asia. Xue et al. (2015) classified Sovia into two species groups based on their distributional patterns as well as on morphological characters. In our phylogeny, the four Sovia species sampled were divided into two unrelated clades (Sovia clade and Clade A), which should be treated as two distinct genera. Consequently, we here recognize the clade with the type species, S. lucasii, along with S. subflava, as the genus Sovia, and the other clade (Clade A) as a new genus (discussed below). The genus Sovia is morphologically distinguishable by the following characters: one large spot across the cell on the forewing, and stigma of male located below the spot in space CuA 1 ; uncus is broad, as broad as tegument, obtuse apically; gnathos elongate, elbow-shaped, densely covered with small spines, and separated; valvae broad, ventrodistal process is irregularly semicircular or rectangular, covered with small teeth at apex or dorsal edge, and centrally with a robust inner, spined, sharp projection. The other five species, S. separata (Moore), S. lii Xue, S. fangi Huang & Wu, S. grahami (Evans) and S. malta Evans also belong to Sovia based on the morphological characteristics (Evans, 1949;Huang, 2003;Huang & Wu, 2003;Xue et al., 2015).

Genera Ampittia Moore, 1881 and Aeromachus de Nicéville, 1890
According to our phylogenetic analyses, the genus Ampittia is a polyphyletic assemblage with the species analysed here separating into two phylogenetically distinct clades. The clade including nana and dalailama clustered with the genus Aeromachus, and the other clade was comprised of three species including the type species, A. dioscorides. The genus Ochus was nested within the second clade. The two species, nana and dalailama, traditionally assigned to Ampittia, share an unbifid uncus of the male genitalia with Aeromachus (the type species Thanaos stigmatus Moore), whereas species of Ampittia (the type species Hesperia dioscorides Fabricius) share bifid uncus (Fig. 4). Therefore, we agree with Wu & Hsu (2017) that these two species should be transfer to Aeromachus, as Aeromachus nana and A. dalailama. The monobasic genus Ochus (the type species Cyclopides subvittatus Moore) shares the following characters with Ampittia: male genitalia with uncus as broad as tegumen and bifid apically (Fig. 4). Based on the phylogenetic analyses and morphological characteristics, Ampittia and Ochus should be merged into one. According to the rule of priority, subvittatus is herewith subsumed under Ampittia, as Ampittia subvittatus comb.n.
Currently, the genus Aeromachus comprises 17 species distributed in the Oriental and Palearctic Regions (Yuan et al., 2015a;Kunte et al., 2017;Wu & Hsu, 2017;present study), but the knowledge on the inter-relationships within Aeromachus is limited. Larsen & Congdon (2012) expected to establish a new genus for the African species of Ampittia. However, they concluded that the African species were congeneric with the Asian species using a combination of biological information and male genitalia characters as well as wing venation. We were unable to sample the African taxa to settle this issue. According to our proposal, Ampittia now consists of three African species and four Asian species, whose inter-relationships are uncertain. Further studies are warranted to investigate the species-level relationships among and within Aeromachus and Ampittia.
Genus Thoressa Swinhoe, 1913 The genus Thoressa traditionally includes 25 species distributed in the Oriental and Palaearctic Regions (Yuan et al., 2015a). Morphologically, these species are complex and diverse. Our phylogenetic analyses clearly showed that the genus is a polyphyletic assemblage. The eight species examined here formed four distinct clades: masoni clade, varia clade, hyrie clade and Clade B (see below for details). The type species, T. masoni, shows a sister relationship with a pair of genera, Parasovia and Sebastonyma, with strong support values. This species differs morphologically from most of the representatives of the traditional concept of Thoressa in the following characters: typical 'Halpe brand' absent; uncus slightly concave distally; tegumen with short and angled shoulders; uncus narrower than tegumen; gnathos small, spoon-shaped and covered with spines, and separated distally; valvae symmetric with a complex and serrate footstalk (Evans, 1949;present study). It does, however, share the following morphological characteristics (Fig. 5): angled shoulders of tegumen, gnathos, symmetric valvae and yellow spots in spaces Rs-CuA 2 on hindwing upperside. Thoressa masoni Moore, along with the unsampled T. decorata Moore, T. monastyrskyi Devyatkin and T. honorei de Nicéville are the only species within the genus (Evans, 1949;Devyatkin, 1996;Tsukiyama et al., 1997). Therefore, we remove all but four species -T. masoni Moore (type species), T. decorata Moore, T. monastyrskyi Devyatkin and T. honorei de Nicéville -from the genus. Among these four species, only T. decorata also shares the male 'Halpe brand' and distally bifid uncus with most of the traditionally recognized Thoressa. However, the above-mentioned morphological similarity is prominent, suggesting that these four species are members of the same genus. Further study is necessary to confirm this hypothesis.
Genus Pedesta Hemming, 1934 Pedesta is a replacement name for Pedestes Watson, which was pre-empted by Pedestes Gray by Hemming (1934). According to Evans (1949), the genus Pedesta morphologically resembles Thoressa, but can be separated from the latter by the following characters: antennal apiculus blunt and obtuse, with nudum of 11 to 12 segments; male genitalia lateral processes absent. Actually, the nudum is often variable within taxa and not a good diagnostic character (de Jong, 1983;Huang & Zhan, 2004). Huang & Zhan (2004) treated Pedesta as a synonym of Thoressa, based on the nudum number, shape of the apiculus, and gradual process of change in lateral processes within these two genera. However, this arrangement is not appropriate because they did not examine both of the type species, P. masuriensis and T. masoni. In our phylogeny, six species of Pedesta including the type species Isoteinon masuriensis Moore and five species of traditionally recognized as Thoressa analysed here formed a strongly supported monophyletic group, which was widely separated from the masoni clade. Our morphological study showed that these species shared the following morphological characteristics: (i) males of most of the species have a stigma in space CuA 2 and 2A on the upper side of the forewing; (ii) gnathos developed, fused with each other, and distally armed with small spines; and (iii) valvae asymmetric (Fig. 6). For lateral processes of tegumen, we found that this taxon was different from that in the traditional concepts of Pedesta or Thoressa, changing gradually within all these species of the Pedesta clade (Fig. 6) as follows: absent (P. masuriensis), short and obtuse (P. baileyi), short and slightly pointed (T. latris), short and pointed (P. gupta, P. fusca), long, slender and pointed (P. submacula). Therefore, the genus Pedesta recognized here includes all the traditionally assigned Pedesta species and some Thoressa species sharing the above morphological characters.

Implications for taxonomy and biogeography
Reconstructing the ancestral range of the tribe Aeromachini clearly pointed to Southeast Asia as its area of origin, from which it has subsequently expanded to the adjacent regions. Within this tribe, the common ancestors of most genera originated in Southeast Asia, which is characterized as in situ diversification, and then subsequently underwent a range expansion to neighbouring areas. Our phylogenetic analyses showed that the genus Arnetta as defined in this study is the basal lineage within the Aeromachini. Biogeographical analyses showed that the most probable ancestral range of Arnetta was in Southeast Asia, which indicated that A. atkinsoni originated in Southeast Asia and subsequently moved eastward and invaded the Indian Subcontinent. Sovia had a similar origination and dispersal pattern to that of A. atkinsoni. A founder-event speciation event was inferred to have occurred for Baracus during the late Miocene, indicating a direction of colonization from Southeast Asia to the Indian Subcontinent. As argued by Holloway (1974), de Jong (1983 and Kunte (2016), these butterflies should have originated after the collision of the Indian tectonic plate with Asia, which is in line with the results of our analyses. In addition, dispersal events of several taxa were dated during the mid-Miocene (c. 14-7 Ma) and coincided with the period when the Indian plate collided with Asia (c. 55 Ma;Hall, 2012). As mentioned above, two species (A. plumbeola and H. lutisquama) colonized the Philippines via long-distance dispersal, both during the mid-Miocene when the proto-Philippines Archipelago was probably connected (Hall, 2002). Three potential colonization routes into the Philippines have been outlined (Wu et al., 2009;Esselstyn & Oliveros, 2010;Heaney et al., 2012;Matuszak et al., 2015;Kyriazis et al., 2018, and references therein). The eastern route was from Borneo to Mindoro using Palawan as a stepping-stone, which may have been a land-bridge during the Pleistocene. The southern route was from Borneo to Mindanao via Sulu Island. A northern route was from mainland Asia and/or Taiwan to the Philippines. The northern hypothesis is only supported by long-distance dispersal, because of the distant isolation of Luzon and the Asian continent by the Bashi strait (Wu et al., 2009). The ancestors of these two species were inferred to have come from Southeast Asia during the mid-Miocene, supporting the northern route. Another long-distance event was also inferred for P. varia dispersing to the Japanese and Sakhalin Archipelago during the mid-Miocene (c. 11.57 Ma). Japan and Sakhalin have been an archipelago since the Miocene (25-15 Ma), when the Japanese Sea separated them from East Asia (Maruyama et al., 1997;Isozaki et al., 2010). The common ancestor of Pedesta likely had colonized Southeast Asia and SWC by the mid-Miocene and diversified there. During the mid-Miocene, the Southwest Mountains (including the Himalayan-Hengduan Mountains) were already high and undergoing extensive uplift (Clift, 2006;Wang et al., 2012). Vicariance thus seems likely to have played a significant role in the diversification of Pedesta. Dramatic crustal deformation induced by the Indo-Asian collision contributed to the complicated landscapes in western China. Massive mountains and deeply carved valleys formed, which acted as barriers to expansion and resulted in speciation, that subsequently gave rise to two interesting lineages, of which one originated from Southeast Asia and another from SWC. Thus, the SWC lineage might have undergone an in situ adaptive radiation due to vicariance. This pattern also can be found in amphibians (Che et al., 2010;Li et al., 2013). On the basis of adult morphological and pupal characters, the African Ampittia species are congeneric to the Asian species, although only the Asian species were sampled in this study. The present Ampittia grouping suggests that there was at least one trans-oceanic dispersal event from Southeast Asia to the Afrotropical Region, although the divergence time of this event was not clear. In summary, our findings indicated that the probable common ancestor of the Aeromachini came from Southeast Asia. Diversification in situ and range expansion to adjacent areas have played essential roles within the Aeromachini, but founder-event speciation and vicariance also have been important factors in shaping the diversification of this tribe. However, this pattern might be challenged if additional African, island, and SWC endemic species samples were included. Evans (1937Evans ( , 1949Evans ( , 1951 treated Hesperiidae of different geographical regions separately. He considered that many genera are distributed both in Africa and Asia, such as Parnara, Pelopidas, Borbo, Isoteinon, Astictopterus, Celaenorrhius, Tagiades, Sarangesa, Caprona, Spialia, Ampittia, Kedestes and Gegenes. Larsen (2005) first challenged Evans's Hesperiidae classification. New taxonomic frameworks of Hesperiidae have recently been proposed (Warren et al., 2008(Warren et al., , 2009Sahoo et al., 2016), although disagreements at the subfamily and tribe levels still exist. Recently, Fan et al. (2016) clarified the phylogeny of Baorini (Hesperiidae: Hesperiinae) and established a new genus for some of the African species traditionally assigned to Borbo. The widespread Borbo was separated into two distinct clades. The clade of four African Borbo (holtzi, perobscura, gemella and an unidentified species) was found to be the basal clade and sister to the rest of the tribe, implying that the probable ancestral area of Baorini may have been Africa. While investigating the phylogeny of the Old World Hesperiidae, we were fortunate to be able to include two African Astictopterus samples in analyses. Surprisingly, the African Astictopterus clade is clearly far separated from the Asian clade, namely from the type species Astictopterus jama (Huang et al., in preparation). Many African and Asia lineages were placed into one genus based on morphological characters, but the molecular phylogeny revealed that these lineages always show relative distant relationships. Examples of this phenomenon also are found in other lepidopteran groups and different zoogeographical regions, including nymphalid butterflies (e.g. Wahlberg & Freitas, 2007;Aduse-Poku et al., 2016) and erebid moths (e.g. Zenker et al., 2017). Such so-called congeneric taxa that occur in different biogeographical regions or on different tectonic plates are not always actually members of the same genus, as we 'traditionally' assumed. Therefore, taxonomic classification of such genera that occur across different biogeographical regions or geotectonic plates needs to be inspected more closely. These taxa, particularly the endemics, are critical in investigating historical biogeography, range expansions and adaptive radiations to shed light on how Earth's biodiversity has been shaped.

Etymology
The name is derived from the genus Halpe with respect to the similarity in external features.

Diagnosis
Antenna almost equal to 2/3 forewing length. Labial palpus erect, densely covered with black scales dorsally and grey-white scales ventrally; the third segment short, porrect, and slightly protruding. Forewing upperside ground colour black; three white spots usually present in space R 3 -R 5 ; two cell spots separated. Underside almost same as upperside, but paler and with a row of obscure or distinct whitish submarginal stripes. Hindwing ground color black, upperside without spots, but underside with black, grey or white striped pattern. Typical 'Halpe' stigma located between spots in discal cell and CuA 1 .

Remarks
We can easily distinguish the new genus from'Sovia' by the following characters: cell spots separated; male stigma located between the spots in space CuA 1 and cell. In Sovia, cell spots conjoined and male stigma located below spot in space CuA 1 . The genus is also similar to Parasovia in wing pattern on the upperside, but markings on the underside of hindwing and male genitalia are completely different, namely socius absent, uncus unbifid and apically flat; whereas Parasovia has long and slender socius, and uncus apically blunt. Based on the combination of characters of two separated cell spots and male genitalia (Evans, 1949;Devyatkin, 1996;Miyazaki & Saito, 2010;Xue et al., 2015)

Etymology
The name is derived from false (Prae-) Thoressa, indicating the previous misplacement.

Diagnosis
Antennae slightly longer than 1/2 forewing length. Labial palpus with second segment stout and erect, covered with ochreous hairs ventrally, third segment black and porrect. Wing underside with submarginal series of yellow spots separated by black veins; forewing with two cell spots or only lower one; hindwing underside usually with three yellowish white spots in spaces Rs, M 3 and CuA 1 . Male stigma below the cell spot.
Male genitalia (Fig. 4f) Tegumen broad, nearly equal to uncus in length with long and slender lateral processes; uncus broad, flat and inwardly hollow, deeply bifid, U-shaped; gnathos well-developed, fused with each other, with spines ventrally; valvae symmetric; costa almost as long as sacculus; footstalk developed with small spines; ventrodistal process rectangular distally, upper edge with a long robust spine, and outer edge straight, covered with small spines; aedeagus straight.

Remarks
The new genus externally resembles Thoressa s.s. and Pedesta, but it can be distinguished from both by the following characters: wing underside with submarginal series of yellow spots; uncus flat and inwardly hollow. Both the molecular phylogeny and morphological characters support this new genus as distinct from all the other known genera. An endemic species from Taiwan, horishana, resembles varia in male genitalia and wing pattern so much that it is sometimes considered a subspecies of varia (Evans, 1949;Shirôzu, 1960;Tsukiyama et al., 1997;Hsu, 2015;Wu & Hsu, 2017). Thus, we transfer this species to Praethoressa, as Praethoressa horishana comb.n. The genus Praethoressa extends its range northward to Sakhalin and the Kuril Islands, across Japan, and southernmost to Taiwan. On the other hand, the distribution of Thoressa (s.s.) (present study) extends further southward than that of Praethoressa, ranging northeastward to Vietnam, and southwestward to southern India. These distributional patterns are in line with our phylogenetic and morphological conclusions.

Supporting Information
Additional supporting information may be found online in the Supporting Information section at the end of the article. Table S1. Information and GenBank accession numbers of the voucher specimens used in this study. *indicates that gene sequence was downloaded from NCBI.
File S1. The best-fit partition schemes and models of molecular evolution of Aeromachini data matrix. The best fitting model was selected by PartitionFinder under BIC. File S2. The models were selected by ModelFinder implemented in IQ-Tree.
File S3. ML tree from RAxML under GTR+GAMMA with best-fit partitions.
File S4. The MCC trees from BEAST. File S5. Taxonomic changes raised by this study; * denotes that the taxonomic changes were only based on morphological characters.