Macromiidae, Needham, 1903

Kosterin, Oleg E., Vierstraete, Andy, Schneider, Thomas, Kompier, Tom, Hu, Fang-Shuo, Everett, Larry, Makbun, Noppadon, Onishko, Vladimir V., Papazian, Michel & Dumont, Henri J., 2025, Molecular phylogenetic analysis of the family Macromiidae (Odonata) worldwide based on a mitochondrial and two nuclear markers, with a short overview of its taxonomic history, Zootaxa 5620 (4), pp. 501-545 : 516-532

Genera of Macromiidae

The genera of Macromiidae were established in 19th century according to the taxonomic concepts of that time which, in particular, overestimated importance of the venation. The chronologically first genus Epophthalmia was described by Burmeister (1839) to include many species currently considered in seven genera of three families. He did not indicate the type species, which was subsequently designated as being E. vittata by Hagen (1867: 62), who cited the personal communication by Burmeister: “Del’ Gattungsname Epophthalmia 1st auf den persönlich ausgesprochenen Wunsch Prof. Burmeister’s seiner E. vittata belassen [The genus name Epophthalmia was retained at the personal request of Prof. Burmeister’s for his E. vittata .]”. Subsequent authors used to include in Epophthalmia in the narrowed sense also some other large species, including some North American ones, all placed in Macromia since the paper by Williamson (1909). The present content of Epophthalmia as including five Asian species has been established after the comprehensive revision by Lieftinck (1931).

The genera Macromia and Didymops were simultaneously established by Rambur (1842). Macromia was proposed for five new species described in the same work, of which only three remained in this genus until now, one was later moved to Epophthalmia and one to Phyllomacromia . The type species was not indicated and was later designated by Kirby (1890). He was aware of the importance of type species of genera and so made such designations for all genera known to him as missing the type species. For this purpose, he usually chose species listed first in the original description of a genus, so he designated Macromia cingulata Rambur, 1842 as the type species of the genus Macromia . This choice appeared to be not so good because this species is more peculiar than ‘typical’ among its congeners for being the smallest, having unusually short and broad hamuli in the male secondary genitalia, and very extensive yellow markings ( Fraser 1936). Moreover, the holotype female (the only specimen which the original description was based on) was mislabelled, even with respect to the continent. Rambur (1842: 138) indicated its origin as follows: “Collection de M. Serville, òu elle est placée parmi les AEschna , et étiquetée sous le nom que je lui ai conservé; indiquée de l’Amérique septentrionale [Collection of M. Serville, where it is placed among Aeschna , and labeled under the name that I have kept for it; indicated from North America]”. Selys (1874) clarified that this was in error and supposed that the true type locality was ‘Bengal’. This species occurs in India, Pakistan and Nepal ( Fraser 1936; Kalkman et al. 2020).

The genus Didymops was proposed for the sole species Didymops servillii Rambur, 1842 , described in the same work ( Rambur 1842). This name was later treated as a junior synonym of Libellula transversa Say, 1840 . Hence, the species under the currently valid name Didymops transversa is the type species of the genus by monotypy as to the original description (the only one mentioned therein).

It is noteworthy that Selys (1871) considered Macromia as the valid name for the genus and subgenus and Didymops as its synonym, even at the subgeneric level. Thus, according to the current ICZN Art 24.2, he acted as the First Reviser who chose the valid name among simultaneously published synonyms. Although seven years later he ( Selys 1878) restored the name Didymops to denote a valid subgenus, this decision persists whenever these two names are considered synonymous.

The name Phyllomacromia was first proposed by Selys Longchamps (1878) for a new subgenus including two African and one Madagascarian species exhibiting foliaceous broadenings of the end of abdomen. Its type species was subsequently designated again by Kirby (1890) to be Macromia trifasciata Rambur, 1842 , most probably since it was chronologically first described. It is the only species of this genus occurring in Madagascar and absent from the African continent. F.C. Fraser, who was the actual author of the text of Tillyard & Fraser (1940), wrote: “several species of Macromia have been described from Africa, but an examination of their genitalia, which I carried out recently, proves them all to belong to Phyllomacromia ” ( Tillyard & Fraser 1940: 390). For some reason, in his later work, Fraser (1954) reclassified the African species under Macromia . However, May (1997) restored Phyllomacromia as a genus including all African and the only Madagascarian species (the type species of the genus) of Macromiidae on the basis of his thorough morphological analysis, first of all concerning the characters of the male vesica spermalis and cerci. This treatment is unequivocally accepted at present.

The genus Azuma was proposed for the only nominal species Macromia elegans Brauer, 1865 , currently Epophthalmia elegans , based on venation characters ( Needham 1904). Later Williamson (1909) found these characters unreliable. Ris (1916) assumed the names Azuma and Epophthalmia as synonyms but some subsequent authors used the former as valid since they considered the latter to be a junior synonym of Cordulia Leach, 1815 , because of its above mentioned too broad sense when it was proposed by Burmeister (1839). At last, the name Azuma was claimed to be a junior synonym of Epophthalmia (as defined by its type species) by Lieftinck (1931). Additionally, the name Azuma turned out to be a junior homonym of a genus of fishes ( Lieftinck 1931).

Species grouping

The large number of species in the genera Macromia and Phyllomacromia and their obvious morphological diversity in their current senses inspired repeated attempts of their grouping. To avoid confusion of the informal species groups with species, further in the text we will designate the species groups with a specific name only, e.g. “ calliope - group” rather than “ M. calliope group”, unless in verbatim citations.

It should be noted that all attempts of grouping in Macromia concerned numerous Asian species but did not involve American species. Laidlaw (1922), based on the marking of the face and mesepisternum and the shape of S10, subdivided the Asian species into three groups: westwoodi -, cincta - and calliope - groups. Fraser (1924) added the fourth, cingulata -group. However, the latter looked unnatural as being too heterogeneous with respect to the male secondary genitalia, as followed from illustrations provided by Fraser (1924; 1936) himself. Lieftinck (1929) added the moorei -group, isolating it from Laidlaw’s calliope -group. The grouping by Lieftinck (1929) was most broadly accepted for a long time.

Thus, up to 1930s, two grouping systems of Macromia had been proposed, of which Fraser’s concerned the Indian species and Lieftinck’s concerned mostly Malesian/Australasian species. Both were broadly accepted for the respective regions but underwent further independent splitting.

It should be noted that Lieftinck considered his groups very informally, not as quasi-taxonomic entities of some more or less definite rank. For instance, in his later paper he mentioned that “ urania clearly belongs to the gerstaeckeri group” ( Lieftinck 1950: 702), without defining that group, and then wrote that “the groups defined earlier were admittedly artificial”. Then he ( Lieftinck 1955) provisionally attributed five North Asiatic species to the amphigena -group (isolating it from his moorei -group), which included Macromia amphigena , but did not provide the diagnosis of this group. At last, Lieftinck (1971) considered “the Papuasian group of Macromia ”, composed of species “very large alike and obviously closely interrelated”, which he further split into three groups: “Group I of M. terpsichore Foerster ”, “Group II of M. melpomene Ris ”, and “Group III of M. chalciope Lieftinck ” ( Lieftinck 1971: 30–32). However, according to his earlier seminal work ( Lieftinck 1929), Macromia terpsichore Förster, 1900 and M. melpomene Ris, 1913 were members of the “I Group of M. westwoodi Selys ”, hence all three of his later Papuasian ‘groups’ were actually subdivisions of his primary westwoodi -group.

However, later authors assumed Lieftinck’s groups rather seriously. The arachnomima -group was independently proposed by Wilson (1993) and Muraki (2010). However, both authors did not consider the fact that Macromia arachnomima Lieftinck, 1953 is very similar to M. aculeata Fraser, 1927 ( Kosterin 2015; Muraki, 2021), which Fraser (1927) had classified under his cingulata -group. Muraki (2004) isolated the urania -group from the calliope - group, although this could be considered the gerstaeckeri -group in the sense by Lieftinck (1950). Muraki (2009) mentioned additional groups, namely daimoji -group and vangviengensis -group, and an unnamed group for an undescribed species, with a reservation that this grouping had not been published yet. Muraki (2014) provided short diagnoses and lists of included species for seven groups, which were the four groups by Lieftinck (1929) plus urania -group, arachnomima -group, and vangviengensis -group by Muraki, but for some reason did not mention his daimoji -group in that paper ( Muraki 2014). Such over-splitting of the SE Asian representatives of the genus in groups seems to be more misleading than helpful.

The Fraser’s grouping system was recently reconsidered by Sadasivan et al. (2023) who, based on the same ‘Fraser’s’ characters of the male cerci and secondary genitalia structures, isolated M. ellisoni from Fraser’s cingulata - group into the ellisoni -group of its own. They also split the Western Ghats species of cingulata -group into two subgroups, flavicincta-bellicosa-irata - and cingulata - subgroups. Isolation of the latter species, peculiar for its small size and short, broad hamuli, looks reasonable.

The grouping within Phyllomacromia spp. was also repeatedly addressed ( Gambles 1979; Legrand 1992; 1993; Dijkstra 2005). Eleven groups were recognised in the latest of these works, with a note that they “strictly serve convenience and (as yet) have no phylogenetic basis, conveying jizz rather than kinship” ( Dijkstra 2005: 23).

Phylogenetic reconstructions

Histone H3–H4 region tree

The phylogenetic tree reconstructed from sequences of the histone H3–H4 region is shown in Fig. 2 . The further text will mention branches with indication of their support as posterior probability values in brackets. All but three sequences of the histone H3–H4 region were obtained in the course of this study, so the set of species in this tree is mostly confined to our species sample.

In this tree, the Macromiidae family is monophyletic with the perfect support (1.00). At the same time the genera Macromidia , Idionyx (both being well-supported monophyletic clusters) and Oxygastra do not form a monophyletic clade, as might be expected. Together with Macromiidae , they form a major clade with a high support (0.92), which does not include C. aenea . In this clade, Oxygastra curtisii ( Dale, 1834) reliably clusters with Macromidia (0.90), while association of their joint cluster with Macromiidae is unsupported (0.53). The genus Idionyx diverges into two clusters, one of which (0.85) includes three species from the Western Ghats of India ( Idionyx corona Fraser, 1921 , I. saffronatus Fraser, 1924 , I. travancorensis Fraser, 1931 ) and I. carinatus Fraser, 1926 from Vietnam, while the other (1.00) contains three species from the eastern part of Oriental Realm ( I. montanus Karsch, 1891 , I. thailandicus Hämäläinen, 1985 , I. selysi Fraser, 1926 ).

The most basal branch of Macromiidae is Didymops , however, the opposed rest of the family is weakly supported (0.68), so it is better to speak of three polytomic clades: Didymops , Macromia and Epophthalmia + Phyllomacromia . The last cluster has the highest possible support (1.00), and inside it, Epophthalmia and Phyllomacromia themselves are also monophyletic with the same support. The monophyly of Macromia is poorly supported (0.68).

Disregarding the weak node of 0.75, the genus Macromia comprises five well supported main clusters, enumerated below from bottom to top.

- The most basal cluster of this genus (1.00) contains M.aculeata and Macromia katae Wilson,1993 corresponding to the arachnomima -group by Wilson (1993) and Muraki (2010).

- One (0.89) of the three other polytomic clusters includes M. pyramidalis and Macromia westwoodi Selys, 1874 , belonging to the westwoodi -group by Lieftinck ( Laidlaw 1922; Asahina 1987a).

- The next cluster (0.94) consists of two subclusters. One of them (1.00) contains nine species (from M. flavocolorata to M. daimoji ), eight from the calliope -group ( Lieftinck 1929; Asahina 1987a; Muraki 2014) and M. daimoji as a sister branch to them, thus being in line with isolation of this species into the daimoji -group of its own by Muraki (2009).

The other subcluster (0.97) contains four species attributed to different groups: M. cupricincta Fraser, 1924 (from the cincta -group ( Lieftinck 1929)), M. irata (from the cingulata -group ( Fraser 1924)), M. viridescens Tillyard, 1911 and M. cydippe Laidlaw, 1922 (from westwoodi -group in the broad sense by Lieftinck (1929; see also 1971)).

- The next cluster (0.91) again consists of two well supported subclusters. One of them (1.00) quite surprisingly includes M. manchurica , M. hamata and M. annaimallaiensis Fraser, 1931 . Macromia manchurica was not classified to any group, while M. annaimallaiensis was attributed to the cincta -group ( Sadasivan et al. 2023). The other subcluster (0.89) includes M. moorei , M. splendens ( Pictet, 1843) , M. malleifera Lieftinck, 1955 and M. unca Wilson, 2004 . Macromia splendens and M. unca were not classified with respect to the groups of Macromia . Macromia moorei belongs to the moorei -group in the sense of Lieftinck (1929). However, later Lieftinck (1955) implicitly subtracted from it his amphigena -group, which included M. malleifera and M. amphigena , of which the former is found in this subcluster while the latter is found in the next cluster.

- The last cluster (1.00) includes two Palaearctic species M. amphigena and M. clio Ris, 1916 of the amphigena group and all five Nearctic species involved, without Palaearctic-Nearctic divergence.

At the low level of the phylogenetic reconstruction of Fig. 2 , some interesting points should be mentioned. The quantitative data on sequence differences are reported below as the number of nucleotide substitutions with the uncorrected p-distances (percentage of variable nucleotide positions) calculated on this base, and the number of indels, if any, but are not shown as alignments for the sake of space.

The sequences of the H3–H4 region of M. flavocolorata from Cambodia and Western Ghats of India (Kerala) do not cluster together, differing in 30 substitutions (uncorrected p-distance 3.5%) and a two-nucleotide-long indel.

The sequences of M. callisto from Cambodia and Thailand also do not cluster together and differ in 45 substitutions (5.2%) and four indels (two of one nucleotide and two of two nucleotides). At the same time, the Cambodian sequence of M. callisto differs from that of M. urania Ris, 1916 only in four substitutions (0.5%) while the Thai sequence of M. callisto differs from that of M. calliope Ris, 1916 with seven substitutions (0.8%) and a two nucleotide long indel. It is noteworthy, however, that the difference between M. calliope and M. flavocolorata from Cambodia is even smaller, consisting of five substitutions (0.6%) and one nucleotide indel. The situation with these three species is complicated and addressed in Discussion.

The sequences of the histone H3–H4 region of M. manchurica and M. hamata differs in four nucleotides only (0.5%). Such small differences would more fit specimens of the same species.

The two subspecies of Macromidia genialis Laidlaw, 1923 (both specimens from Cambodia) appeared very close; they differ in 13 nucleotide substitutions (1.5%) and one nucleotide long indel.

In general, the tree based on the H3–H4 region ( Fig. 2 ) corresponds rather well to the current taxonomical concept. However, in the genus Macromia it supports monophyly of the calliope -group and arachnomima -group while representatives of other involved groups were found in different clusters and subclusters. The association of M. annaimallaiensis from the Western Ghats and M. manchurica from East Asia was not expected.

Of the two alternative treatments of Macromidia , Idionyx and Oxygastra , as genera incertae sedis ( Dijkstra et al. 2013; Paulson et al. 2025) or members of the family Synthemistidae in the broad sense ( Carle et al. 2015), our tree ( Fig. 2 ) supports the former, since these genera do not cluster together.

ITS region tree

The Bayesian tree based on sequences of the other nuclear marker, the ITS region, is presented in Fig. 3 . In general, it is rather similar to the H3–H4 region tree ( Fig. 2 ), but involves more sequences, including those from GenBank.

Again, the family Macromiidae is monophyletic with the highest support (1.00) while Macromidia , Idionyx (both again supported at 1.00) and Oxygastra do not form a monophyletic clade. Oxygastra is recovered as a branch of its own. Curiously, Idionyx , Oxygastra , Cordulia and Macromiidae appeared to comprise a monophyletic clade with the highest support of 1.00 (with Idionyx being a sister clade to the rest), while Macromidia is the most basal clade of the tree. The inner structure of Idionyx is reshuffled as compared to the H3–H4 region tree ( Fig. 2 ).

The ITS tree ( Fig. 3 ) corresponds less to the current taxonomy than the H3–H4 region tree ( Fig. 2 ). First, Didymops is no longer the most basal clade of Macromiidae but is found deep inside Macromia , clustering with M. moorei , M. splendens and M. malleifera . The most basal clade of Macromiidae in the ITS tree (0.97) is now formed by M. arachnomima , M. aculeata , M. katae and M. westwoodi . This clade is opposed to the rest of the family, which is poorly supported (0.76). Hence the genus Macromia has lost its monophyly.

Epophthalmia + Phyllomacromia (1.00) form a sister clade to the large clade (0.97) including the rest of Macromia , without the three above mentioned species but with Didymops . Monophyly of Epophthalmia and Phyllomacromia has the maximum support.

In the remaining Macromia spp. , ten well supported clusters (0.96–1.00) and solitary lineages are resolved:

- Macromia pyramidalis .

- Macromia unca (one sequence).

- Macromia viridescens (one sequence).

- Macromia cupricincta , Macromia cincta Rambur, 1842 and M. irata .

- Macromia cydippe (one sequence).

- The calliope -group: M. calliope , M. callisto , M. flavocolorata , Macromia septima Martin, 1904 , M. urania , Macromia gerstaeckeri Krüger, 1899 (but see ‘Discussion’ concerning this sequence), Macromia chaiyaphumensis Hämäläinen, 1986 , Macromia sp3 .; again together with M. daimoji forming a separate subcluster.

- Macromia moorei , M. splendens , M. malleifera and D. transversa .

- Macromia manchurica and M. annalmallaiensis .

- Five American species of Macromia .

- Macromia amphigena , M. kubokaiya Asahina, 1964 , M. clio .

We see these clusters to be mostly the same as in the H3–H4 tree ( Fig. 2 ), with the following differences:

- The amphigena -group is now only very weakly coupled (0.65) with the American branch.

- Macromia unca gets decoupled from the moorei -group, which is, however, updated with D. transversa .

- Macromia westwoodi is decoupled from M. pyramidalis and moved to the common clade with M. arachnomima , M. aculeata and M. katae .

- Macromia cupricincta and M. cydippe are decoupled from M. cingulata and M. irata and are separate lineages.

In general, the clusters inside Macromia remained mostly the same as in Fig. 2 , while the basal topology of Macromiidae changed substantially.

At the low level, the following notes are necessary.

The sequence of M. flavocolorata from the Western Ghats (Kerala) again does not cluster with those from Vietnam and Cambodia at all, differing from them in a 36 nucleotide long insertion and 24 nucleotide substitutions (4.2%).

FIGURE 3. (Continued).

FIGURE 3. (Continued).

The sequences identified as M. callisto are now found in three different places of the calliope -group cluster. That from Cambodia is again found among M. urania . Its sequence is identical to those of M. urania from Japan and Vietnam, except for a small indel region where it has two nucleotides while the two latter have six and none, respectively. M. callisto from Thailand again clusters with M. flavocolorata and M. calliope and has only one nucleotide different (0.1%) (plus one heterozygous position) from the sequence of M. calliope from Vietnam. The GenBank sequence M. callisto from Malaysia lies separately from other sequences identified as that species.

In the same calliope -group cluster, there are two more pairs of species raising suspicions of their identity or misidentifications.

The sequences of M. calliope differ from those of M. flavocolorata from Vietnam and Cambodia in as few as three substitutions (0.5%) (plus one heterozygous position), the latter two differing from each other in two indels, one and two nucleotides long.

The sequence of M. gerstaeckeri from GenBank and the two sequences of M. chaiyaphumensis are identical.

The sequences of M. daimoji of different origin tightly cluster together without geographic regularity. Their alignment (not shown) exhibits some substitutions and six sites of indels from one- to six-nucleotide-long: five of them are in the sequence from Taiwan and one in that from Russia; one of the Japanese sequences also differs from the two others in a two-nucleotide-long indel in one of the same sites.

The four sequences of M. cupricincta are identical, while the two sequences of M. berlandi Lieftinck, 1941 differ from them only in one substitution (0.2%) and a one-nucleotide-long deletion.

In all three cases where subspecies of the same species were involved, their sequences appeared to differ rather significantly, as follows.

Macromia amphigena amphigena , M. amphigena fraenata , M. clio and M. kubokaiya are very close in the phylogenetic reconstructions of Fig. 3 but nevertheless formed four different and very well supported (0.94-1.00) clusters, of which the two presumed subspecies of M. amphigena are most distant. Their sequences appeared to differ more in indels than in substitutions. Only 19 nucleotide positions are variable, but seven regions were affected by indels (two of them one-nucleotide-long). There are six (0.9%) diagnostic (found only in this taxon) substitutions and a diagnostic eight-nucleotide-long insertion of M. kubokaiya , four (0.6%) diagnostic substitutions and a diagnostic eight-nucleotide-long insertion of M. amphigena amphigena , one (0.2%) diagnostic substitution and two diagnostic two-nucleotide-long deletions of M. amphigena fraenata , and just one (0.2%) diagnostic substitution of M. clio , the latter species being variable for longer indels. The three sequences of M. amphigena fraenata contain six (0.9%) variable positions and two indels (one- and two-nucleotide long) without geographic regularity.

Macromia moorei moorei from Himachal Pradesh, India, is found inside the tight cluster formed by M. moorei malayana . The sequence of the former has a four-nucleotide-long insertion and four nucleotide substitutions (0.6%) not found in those of the latter, which in turn differ from each other in few substitutions.

The difference between the two subspecies of Macromidia genialis appeared substantial. The sequence of M. genialis buusraaensis Kosterin, 2018 differs from that of M. genialis shanensis Fraser, 1927 in two eight-nucleotide-long insertions, a one-nucleotide-long insertion, three one-nucleotide-long deletions, and 12 nucleotide substitutions (1.5%).

All the five sequences of E. vittata regardless of their origin ( India, Cambodia, Laos and Vietnam) are identical.

COI tree

The phylogenetic tree reconstructed on the base of the mitochondrial COI barcoding fragment is shown in Fig. 4 .

In addition to Macromidia , Idionyx and Oxygastra , the genera of the ‘GSI clade’ in the sense by Ware et al. (2007) are now updated with Synthemis . They again do not form a monophyletic clade. As in the ITS-tree ( Fig. 3 ) and unlike the H3–H4-tree ( Fig. 2 ), Macromidia occupies the most basal position while Idionyx clusters with Macromiidae . However, Macromidia donaldi ( Fraser, 1924) appeared outside the main Macromidia clade and comprises an independent branch, like Cordulia aenea and Synthemis eustalacta ( Fig. 4 ). Oxygastra now shows a tight clustering with Idionyx ( Fig. 4 ), although it clustered with Macromidia in the H3–H4 tree ( Fig. 2 ) and was independent in the ITS-tree ( Fig. 3 ).

As in the two other trees, Macromiidae are again monophyletic with the maximum support (1.00) in the COI tree ( Fig. 4 ). Inside it, the most basal clade (1.00), opposed to the rest of the family, is represented by M. arachnomima , M. aculeata and M. katae . The rest of the family (0.98) then diverges into four clades with moderate support:

- Phyllomacromia (0.84).

- Epophthalmia (0.98).

- The calliope -group without M. daimoji (0.85).

- The rest of Macromia with D. transversa (0.90).

The following smaller clusters and separate lineages were revealed with supports above 0.8:

- Macromia daimoji (1.00).

- Macromia westwoodi and M. cydippe (0.95).

- Macromia cupricincta , M. irata , M. cincta and M. viridescens (0.86).

- Macromia pyramidalis and M. unca (0.85).

- Macromia annaimallaiensis .

- Macromia moorei , M. malleifera , M. splendens (0.99).

- Macromia manchurica and M. hamata (1.00) (just four variable positions, 0.6%, without any regularity of substitutions as to the three sequences).

- Disymops transversa (1.00).

- The American species of Macromia (0.93).

- Macromia amphigena sspp and M. clio (1.00).

We see that

- clustering of Phyllomacromia with Epophthalmia is no longer supported (0.53);

- Didymops is found inside Macromia , as in the ITS tree ( Fig. 3 );

- Macromia daimoji is now decoupled from the calliope -group;

FIGURE 4. (Continued).

FIGURE 4. (Continued).

- Macromia annaimallaiensis is decoupled from M. manchurica , which it clustered with in other trees ( Figs 2–3 ).

In general, the COI tree ( Fig. 4 ) is more or less similar to the H3–H4 ( Fig. 2 ) and ITS ( Fig. 3 ) trees as having similar clusters, but they are less supported and the position of some species differed strongly, especially of M. donaldi ( Fraser, 1924) , O. curtisii and M. annaimallaiensis .

At the species level the following details are the same as in the ITS tree:

- Macromia moorei moorei ( India) is again found inside the branch of M. moorei malayana . The haplotypic network constructed for their COI sequences ( Fig. 5 ) did not reveal the two subspecies but shows that the sequence from India, Vietnam and the two Thai sequences taken together are equidistant.

- Macromia hamata is inside the branch of M. manchurica .

- Maccromia amphigena fraenata clusters with M. clio rather than with M. amphigena amphigena . To better illustrate the latter case, we constructed the haplotypic network for the COI sequences of M. amphigena sspp. and M. clio ( Fig. 6 ). The clusters of the three taxa included are separated by significant distances (21 and 44 mutation steps; 3.2 and 6.3%), so best corresponding to closely related but distinct species. The sequences of M. amphigena fraenata from West Siberia and Primorye differ in six (0.9%) substitutions only.

The situation in the calliope -group in the COI tree ( Fig. 1 ) is as complicated as in the two previous trees ( Figs 2– 3 ) (unfortunately, we did not manage to get the ITS sequence from the specimen of M. calliope itself), as follows:

- The sequences of M. daimoji from Korea, Russia and Taiwan (the latter woud be identified as M. chui in recent past) are very similar: just four (0.6%) substitutions in total, three of which are in the sequence from Taiwan.

- Macromia flavocolorata from the Western Ghats (Kerala) again does not cluster with those from Vietnam and Cambodia but lies separately.

- The sequences of M. callisto are found in two different places, again near M. flavocolorata from Vietnam and Cambodia and near M. urania , to which M. chaiyaphumensis is now added.

To illustrate the situation in the calliope -group, we constructed its COI haplotypic network ( Fig. 7 ). The sequence of M. chaiyaphumensis differs from the closest ones of M. urania and M. callisto from Cambodia in four mutation steps only (0.6%). Macromia sp1 , Macromia sp3 and M. flavocolorata from Kerala are species most distant from the rest.

Two sequences of Macromidia genialis shanensis from Thailand and Cambodia and the sequence of M. genialis buusraaensis from Cambodia appeared equally distant from each other ( Fig. 8 ), having in total 101 (as much as 14.5%) variable positions.

The sequences of E. vittata contained 18 variable positions (2.5%) showing no geographical regularity, but ten substitutions were confined to the beginning of the sequence of a specimen from Cambodia.

Joint analysis

In order to summarise and ‘average’ the phylogenetic signals from the three markers involved, we undertook their joint analysis using StarBeast. This software (i) is able of utilising any number of sequences and markers available for each species and (ii) takes into account that sequences evolve not alone but as incorporated in some species, which actually evolve, and reconstructs the most plausible phylogenetic tree of species rather than sequences. To make this possible, it takes into account the species identifications of each particular input sequence and hence is influenced by expert opinion of identifiers. Because of this, we removed from this analysis all sequences identified as M. callisto , since this identification appeared confusing in view of the phylogenetic reconstructions of particular markers ( Figs 2–4 ). We also removed the only sequence of M. gerstaeckeri (of ITS region), which was obviously misidentified M. chaiyaphumensis (see ‘Discussion’). Also, we separated M. flavocolorata from India, Kerala as a species different from M. flavocolorata from Vietnam and Cambodia, as was obvious from the above analysis. The resulting tree is shown in Fig. 9 .

In this tree, the family Macromiidae is monophyletic with a high support (0.99), as in all other trees. Curiously, the genera Idionyx , Oxygastra , Macromidia and Cordulia cluster with it to form a major clade (0.96), but do not reliably cluster with each other, while Synthemis is left outside, but this could be an artifact due to representation of the latter with just one COI sequence. The genus Macromia becomes a monophyletic but rather not supported (0.65) major clade and only if Didymops is included in it.

As could be expected from a tree based on three rather contradicting markers at once, most of its clades are weakly supported. However, the remaining highly supported clades should correspond to actually existing monophyletic entities and worth being enumerated. These are:

- The genus Idionyx (0.98), divided into three inner branches, two of which the have maximum support (1.00): I. saffronatus with I. travancorensis and four species ( I. montanus , Idionyx murcia Lieftinck, 1971 , I. selysi , I. thailandicus ) referring to the yolanda -group.

- The genera Phyllomacromia and Epophthalmia (both supported at 1.00), together also forming a very well supported (0.96) clade.

- The calliope -group of Macromia (0.99), including M. daimoji as its most basal branch.

- Macromia arachnomima , M. aculeata and M. katae (1.00).

- Macromia cupricincta , M. berlandi , M. cincta and M. irata (1.00).

- Macromia splendens , M. moorei , M. malleifera together with D. transversa (0.93)!

-A large clade with the support of 0.99 with three inner branches of the highest support of 1.00: (i) M.manchurica , M. hamata and M. annaimallaiensis ; (ii) the North American species, and (iii) the amphigena -group.

We also reconstructed the StarBeast tree only for our two nuclear markers, the histone H3–H4 and ITS regions. It showed the same, locally a bit less supported results for Macromiidae , so we do not show it. The pattern of the ‘GSI clade’ genera was somewhat different, that could be expected from the particular trees of the nuclear markers ( Figs 2–3 ).

Figures

FIGURE 1. Some representatives of the family Macromiidae (under the names according to this paper) photographed in nature: a—Macromia daimoji, Taiwan; b—M. urania, Taiwan; c—M. clio, Taiwan; d—M. fraenata—Russia, Novosibirsk Province; e—M. manchurica, Russia, Primorye; f—M. annaimallaiensis, India, Kerala, Idukki District; g—M. miniata, India, Kerala, Idukki District; h—Epophthamia vittigera, Thailand, Phang Nga Province; i—E. elegans, Taiwan; j—E. vittata, Cambodia, Ratanakiri Province; k—Phyllomacromia picta, South Africa, Marloth; l—P. contumax, South Africa, Marloth; a–e, i–l—males; f, h—females; g—copula. Not to scale.

FIGURE 2. Phylogenetic tree (phylogram) of representatives of the family Macromiidae and some related genera reconstructed with the MrBayes 3.2.7 software from sequences of the nuclear histone H3–H4 region (partial sequences of the genes coding for the core histones H3 and H4 and the non-coding spacer between them). Posterior probabilities above 0.5 are indicated at respective nodes. Specimen numbers are indicated for sequences obtained in this study and accession numbers are indicated for the three sequences reconstructed from SRA archives.

FIGURE 3. Phylogenetic tree (phylogram) of representatives of the family Macromiidae and some related genera reconstructed with the MrBayes 3.2.7 software from sequences of the nuclear ITS region, containing internal transcribed spacers ITS1 and ITS2 and the 5.8S rRNA gene between them. Posterior probabilities above 0.5 are indicated at respective nodes. Specimen numbers are indicated for sequences obtained in this study and accession numbers are indicated for sequenced adopted from GenBank.

FIGURE 4. Phylogenetic tree (phylogram) of representatives of the family Macromiidae and some related genera reconstructed with the MrBayes 3.2.7 software from sequences of the barcoding fragment of the mitochondrial cytochrome oxidase I (COI) gene. Designations as in Fig. 3.

FIGURE 5. Haplotypic network of the analysed COI gene fragment of the available sequences of Macromia moorei ssp. constructed with POPART 1.7 software. Specimen numbers are indicated for sequences obtained in this study and accession numbers are indicated for sequenced adopted from GenBank.

FIGURE 6. Haplotypic network of the analysed COI gene fragment of the available sequences of Macromia amphigena, M. clio and M. kubokaiya. constructed with POPART 1.7 software. Designations as in Fig. 5

FIGURE 7. Haplotypic network of the analysed COI gene fragment of the available sequences of Macromia spp. of the calliope-group in the broad sense constructed with POPART 1.7 software. Designations as in Fig. 5.

FIGURE 8. Haplotypic network of the analysed COI gene fragment of the available sequences of Macromidia genialis ssp. constructed with POPART 1.7 software. Designations as in Fig. 5.

FIGURE 9. Phylogenetic tree (phylogram) of species reconstructed with the StarBeast software using the Bayesian approach based on all available sequences of the three markers involved in this study. Designations as in Fig. 3.