Phylogeny of the Synlestidae (Odonata: Zygoptera), with an emphasis on Chlorolestes Selys and Ecchlorolestes Barnard

The Synlestidae (Odonata: Zygoptera) of southern Africa comprise some highly localized species. All but one species are endemic to South Africa, and many to the Cape Floristic Region. Here we present the first phylogenetic reconstruction of the southern African Synlestidae using nuclear and mitochondrial molecular data. The genera Ecchlorolestes and Chlorolestes are monophyletic, and we propose that the Neotropical family Perilestidae consisting of two genera, Perilestes and Perissolestes, be sunk within Synlestidae. We discuss the intra-familial relationships for the southern African Synlestidae.

Alignment. Initial sequence alignments were made using Clustal-X 8 and the resulting 28S files were then aligned manually in Microsoft Word using secondary structural models. phylogenetic reconstruction. Data were partitioned and analyzed using IQTREE 2 (Trifinopoulos et al., 2016), and separate gene trees were also reconstructed. For the combined dataset, a GTR + F + R3 model 9 was implemented; for the COI only tree a TIM2 + F + G4 model and for the 28S only tree a TN + F + R2 model were implemented.
Biogeographical analyses. We evaluated geographical patterns in Synlestidae using the parsimony ancestral state reconstruction function in Mesquite 10 . We assigned the taxa in our phylogeny to one or more of three biogeographical regions: Southern, New World, Central Africa and Australasia. We ran these analyses on our consensus maximum likelihood tree. We set no constraints, and taxa were considered to have equal ability to disperse to each area.
Morphology. We examined the morphological features of several adult Synlestidae (Ecchlorolestes, Chlorolestes, Nubiolestes Fraser, 1945) using standard stereo microscopy, to assess synapomorphies among and within the genera Ecchlorolestes and Chlorolestes. We describe these traits below. We evaluated trait evolution patterns using the parsimony ancestral state reconstruction function in Mesquite 10 . We ran these analyses on our consensus maximum likelihood tree. In Ecchlorolestes, a sclerotized medial spine on the ligula is absent (Fig. 3E) as in Perilestes (Fig. 3A); in Chlorolestes, C. apricans, C. conspicuus and C. umbratus possess a scleritized medial spine shaped like a scimitar blade with a hollow canal/channel at its tip (Fig. 3C), perhaps for sperm transfer. Chlorolestes draconicus, C. elegans, C. fasciatus and C. tessellatus have a flexible medial spine on the genital ligula with a flap that covers the tip of the ligula, much like a pitcher plant (Fig. 3D). This is perhaps similar to the Nubiolestes form with a sclerotized medial spine tip in the form of a funnel on the genital ligula (Fig. 3B).
Kennedy 11 examined C. fasciatus and C. tessellatus and described a hood-like structure present covering the penis tip in the secondary genitalia of Euchlorolestes (synonymised as Chlorolestes by Barnard 12 ). This flap-like structure is similar to that of Nubiolestes (Fig. 3B), whose ligula ends in a narrow tip, shaped like a partially flattened funnel.

Results
Synlestidae s.s. was recovered with 77% bootstrap support. The earliest branching lineages in the topology were Synlestes and Episynlestes. Ecchlorolestes and Chlorolestes were recovered as a clade (69%), with the inclusion of Nubiolestes, Megalestes and Phylolestes (Fig. 7). The NCBI sequence of E. nylephtha was of 28S only ( Fig. 8), which may explain why this sequence did not fall within the clade comprising the remaining members of this species. Phylolestes was recovered as sister to Nubiolestes + Chlorolestes (68%). Within Chlorolestes, C. umbratus was recovered as sister to the remaining Chlorolestes (84%). C. apricans was recovered as sister to (C. conspicuus (C. elegans (C. fasciatus + C. tessellatus))) (73%). The NCBI sequence for C. tessellatus was 28S only (Fig. 8), which may explain why it did not group with the remaining C. fasciatus. www.nature.com/scientificreports/ The median spine located on the ligula of the male secondary genitalia seems to have evolved once, as did a tooth on the caudal appendages. The hood-like structure on the male ligula and evenly spaced concavities between ovipositor teeth were synapomorphies for the grouping of C. elegans, C. fasciatus, C. tessellatus (and C. draconicus, although not sequenced here). The presence of a basal spine on the cercus, an ovipositor with small teeth equidistant from one another, and a medial spine on the ligula, with RP 3 originating before the subnodus were synapomorphies for both species of Ecchlorolestes. www.nature.com/scientificreports/ Biogeographical analyses using the parsimony ancestral state reconstruction function in Mesquite suggest a single African origin of Synlestidae (Fig. 9). Megalestes, recovered as a likely sister to the remaining Synlestidae, occurs in the Oriental Region, and our topology suggests that Phylolestes has an independent bioegeographical origin into the New World, most likely due to dispersal into the Caribbean.

Discussion
Biogeography. The present-day distributions of the Synlestidae-and Odonata in general-reflect millions of years of geographic isolation of South Africa from other areas of the continent, and high-mountain building in the south-west in particular, coupled with a lack of glaciation events for over 200 million years. This has contributed to considerable speciation and endemism 6,17 , particularly in the Cape Floristic Region (CFR), but also in the high KwaZulu-Natal Drakensberg mountains 6 . Although some individuals of Synlestidae species do occur at low www.nature.com/scientificreports/ elevations (e.g. C. fasciatus at near sea level at Mtamvuna), most representatives of the family commonly occur at higher elevations 18 , with higher elevations being common for Neotropical Phylolestes. Perilestidae, however, often occur at low elevations in the Amazon basin. In a warming global climate, the preference for high elevations may leave some montane specialists, such as C. draconicus without suitable habitat 6 . Isolated populations, such as that of C. tessellatus in Sevenweekspoort, at least a hundred kilometers from the nearest population, with arid and unsuitable habitat in between 17 , may face a similar isolationist situation. Ecchlorolestes was first erected as a genus in 1937 by Barnard 12 . In 1962, Pinhey 19 suggested that at least E. nylephtha should be considered as a member of Chlorolestes, but this was reversed 20 . For example, the two species in the genus are unlike Chlorolestes in their possession of a distinct basal tooth on the superior anal appendages 16 . The tooth is stubby in E. peringueyi (Fig. 4L) and narrower in E. nylephtha (see 16 for figures of anal appendages, and Fig. 4K here). However, other closely related taxa also have a similar tooth to Ecchlorolestes, such as the Australian Synlestes weyersii, whose teeth on the superior appendages greatly resemble E. peringueyi. In the Australian lestid Austrolestes cingulatus (Burmeister), the teeth are similar, although each tooth is much shorter and stubbier than E. peringueyi. Perhaps the presence of a tooth on the superior appendage has simply been lost and gained multiple times, although this requires testing with a thorough morphological evaluation, as animal genitalia evolve rapidly with respect to other morphological traits due to sexual selection 21 . The differences in the ovipositor armature between Chlorolestes (robust teeth, Fig. 5C-I) and Ecchlorolestes (teeth small, Fig. 6) may be an adaptation to laying in soft tissue (lichens for E. peringueyi, and ferns/mosses for E. nylephtha), in comparison with harder tissue for the southern African Chlorolestes (terminal twiglets of sclerophyllous vegetation). A phylogenetic hypothesis based on fossil and extant zygopteran taxa would provide further insight into the evolution of synlestid morphology.
In general, Chlorolestinae have very petiolate wings, and occur in association with montane streams, especially deposition pools. Their geological history is not well known, but there are fossils of putative Synlestidae, such as Eolestes 22-24 from the Eocene (roughly 56-34 million years ago) and Gaurimacia sophiae 25 from the Late Jurassic, Early Cretaceous (roughly 145 mya). extant Synlestidae in the new world. There is but one New World representative of Synlestidae, Phylolestes in the Caribbean: is this taxon there due to dispersal? The fossil Eolestes described by Cockerell 22 and considered by Nel and Paicheler 23 to be a putative Synlestidae, was recovered from the New World, in the United States of America. Perhaps Synlestidae species were once more widespread, but have since gone extinct, except for representatives in the Caribbean, Australasia, and southern Africa. If Perilestidae are indeed Synlestidae, as our topology and morphological data suggest, then the Neotropical distribution of Phylolestes may instead reflect an ancestral range that spanned Gondwana. The geological age of Hispaniola is Late Cretaceous to early Cenozoic 26 . Perhaps Phylolestes colonized or speciated on this island, while other Neotropical members of the family went extinct. An autapomorphy for Synlestidae + Perilestidae is the strongly arched CuP (Fig. 2) at its base where it meets the extremity of the quadrangle. With this revised status, New World Synlestidae are then comparatively the most species rich with 21 species compared with taxa from Asia (19), Africa (9) and Australia (7). Do larval characteristics support or refute uniting Perilestidae and Synlestidae? We considered all larval descriptions of perilestids [27][28][29][30] and synlestids known to us [31][32][33][34][35][36] . The caudal lamellae in all genera are very similar in all genera, but the larva of Megalestes (Synlestidae) differs more from all of the other genera (Perilestes, Perissolestes, Phylolestes, Chorismagrion, Episynlestes, Synlestes and Nubiolestes) referenced above. The shape of the www.nature.com/scientificreports/ labium and delicate spider-like legs in Megalestes differ from all of the other genera listed above. In summary, the differences among genera within Synlestidae, based on the published literature, seem to be at least as great, or greater, than differences between Perilestidae and Synlestidae. Any argument to sustain Perilestidae as separate from Synlestidae based on larvae does not seem to be particularly significant. Arguments as to differing altitude preferences also do not seem to be valid reasons for maintaining Perilestidae as separate from Synlestidae, as these can vary even with a single genus. For example, C. draconicus only occurs at high elevations, but many others either occur in a wide elevational range, (C. fasciatus) or even at low elevations (C. tessellatus, C. conspicuus). It is therefore unlikely that elevation provides a valid reason for maintaining the two families as separate.
taxonomy and the status of the southern African genera in Synlestidae. Based on our results, Ecchlorolestes and Chlorolestes are valid taxonomic groups, as also suggested by larval morphology 34 . Some subgroups within Chlorolestes have been suggested based on wing banding. Chlorolestes umbratus, C. tessellatus, C.  www.nature.com/scientificreports/ fasciatus, C. elegans possess banded wings in all or some populations. Our topology suggests that this banding is perhaps not so much a reflection of evolutionary history as localized selection pressure. For example, wing banding in C. tessellatus populations varies according to geographical area, with some populations in the Eastern Cape with heavily banded wings, and others, for example in KwaZulu-Natal with no banding at all. One population, also in KwaZulu-Natal, has very weakly banded wings 37 . Similar polymorphism in wing maculation occurs in Sinolestes editus Needham and some species of Orolestes McLachlan, 1895, which belong to the family Lestidae) 38,39 . Here, we propose sinking the family Perilestidae within the Synlestidae based on (a) their position as sister to the large clade containing Nubiolestes and the Synlestidae, (b) morphological characters, such as the strongly arched CuP at its base where it meets the extremity of the quadrangle and ovipositor, described above.

conclusions
Ecchlorolestes and Chlorolestes are monophyletic. We suggest that given the molecular topology and morphological data, Perilestidae should be considered members of the Synlestidae. Phylolestes is sister to Chlorolestes, suggesting that there was an African origin of the clade containing Nubiolestes, Ecchlorolestes, Phylolestes and Chlorolestes, and subsequent dispersal to the Caribbean by Phylolestes, unless Perilestidae are also included within the family, in which case perhaps a broader origin is possible, as supported by the fossil evidence and Mesquite analysis.